Science Prediction Universe (SPU) public tasks

.gitattributes

@@ -0,0 +1,62 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text
+tasks/hydrogel-additional-formulation-adhesion/tests/answers.json filter=lfs diff=lfs merge=lfs -text
+paper/Science_Prediction_Universe.pdf filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,132 @@
+---
+license: cc-by-4.0
+pretty_name: Science Prediction Universe
+language:
+- en
+size_categories:
+- n<1K
+tags:
+- science
+- benchmark
+- agent-evaluation
+- harbor
+- chemistry
+- materials-science
+- biology
+- medicine
+configs:
+- config_name: default
+  data_files:
+  - split: test
+    path: task_cards.parquet
+---
+
+<p align="center"><img src="assets/spu-banner.webp" alt="Science Prediction Universe #Benchmark, UniPat AI" width="100%"></p>
+
+# Science Prediction Universe (SPU)
+
+[Paper](paper/Science_Prediction_Universe.pdf) · [Homepage](https://spu.unipat.ai/) · [Blog](https://unipat.ai/blog/SciencePredictionUniverse) · [GitHub](https://github.com/UniPat-AI/SPU)
+
+<p align="center"><img src="assets/spu-overview.png" alt="An SPU task: the experimental definition goes to a wet lab and to an LLM agent; the agent's predicted ranking is scored against the hidden measured ranking by Spearman correlation." width="900"></p>
+
+Science Prediction Universe (SPU) is a benchmark of hidden wet-lab outcomes. Each task fixes the candidates, materials, conditions and measurement protocol of a real experiment, and the agent ranks the candidates by the measured outcome. Rankings are scored by Spearman correlation against the measurements.
+
+This repository contains the **100 public tasks**, built from 85 papers published online between January 2025 and August 2026 in 11 Nature Portfolio journals: 39 chemistry, 29 materials science, 20 biology and 12 medicine tasks in 29 subfields. The 16 private tasks of SPU come from unpublished wet-lab experiments and are not released.
+
+**The reference answers are included.** Every task ships with its measurements, grader and oracle solution, so anyone can run and score it. These tasks are therefore not a held-out test set.
+
+<p align="center"><img src="assets/coverage.png" alt="Coverage of SPU: the 116 tasks by discipline and subfield, their source journals and the number of experiments per task." width="820"></p>
+
+## Results in the paper
+
+Thirteen agent configurations, each run once on every task with an eight-hour budget, reach a best mean below 0.5 on a scale where a perfect ranking scores 1 and random rankings average 0.
+
+<p align="center"><img src="assets/results.png" alt="Thirteen agent configurations ranked by mean Spearman score on all scored tasks, with 95% bootstrap intervals." width="820"></p>
+
+Means over the public and private tasks on which each configuration has a scored run, with 95% bootstrap intervals over tasks. Hatched bars are means on the 28 tasks that all thirteen configurations completed, and the right-hand column ranks the configurations on those tasks. See the paper for task-adjusted means, paired tests and cost.
+
+## Contents
+
+| Path | Contents |
+| --- | --- |
+| `task_cards.parquet` | One row per task: `task_id`, `title`, `summary`, `category`, `subdomain`, `task_path` (the `default` config, `test` split) |
+| `tasks/<task_id>/` | A [Harbor](https://harborframework.com/) task: `instruction.md`, `task.toml`, `environment/` (solver image and public inputs), `tests/` (grader, measurements and answers), `solution/` (oracle) |
+| `SOURCES.json` | Source paper of each task: DOI, title, authors, journal and URL |
+| `SHA256SUMS` | SHA-256 of every file |
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+cards = load_dataset("UnipatAI/SPU", split="test")
+```
+
+Run a task with Harbor 0.25.0 (Python 3.12 or later) and Docker. This oracle check replays the reference answer and needs no GPU:
+
+```bash
+hf download UnipatAI/SPU --repo-type dataset --local-dir SPU && cd SPU
+harbor run -p tasks/fluorene-dehydration-concentrations -a oracle -e docker \
+  --override-gpus 0 --override-cpus 1 --override-memory-mb 1024 -o ../jobs
+```
+
+In the paper, each run had 8 CPU cores, 16 GiB of memory, one NVIDIA H20 GPU, web access and an eight-hour budget, as set in each `task.toml`. The grader runs in a separate container without network access. The solver should see only `instruction.md` and `environment/public/`; keep `tests/`, `solution/` and `SOURCES.json` out of its reach.
+
+## Scoring
+
+Each task has one or more comparison groups. The grader computes the Spearman correlation between the submitted ranking and the measured values in each group and combines the groups with the fixed weights given in `instruction.md`. A perfect ranking scores 1 and random rankings average 0. Tied measurements get average ranks, so on 33 tasks the best possible score is slightly below 1. A missing or malformed submission scores −1, the same as an exactly reversed ranking: check `submission_status` in the grader's `report.json`.
+
+The paper's headline metric, the task-adjusted mean, is fitted to the scores of all thirteen evaluated configurations and cannot be computed from a single run. When reporting results on these tasks, give the mean over scored tasks, the number of scored tasks, the model version, reasoning effort and harness. The paper's audited web tools and source-identifier screening are not part of this release, so results obtained without them are not directly comparable to the paper's.
+
+## Sources
+
+Each task is built from one published paper, cited in `SOURCES.json`. Some biology and medicine tasks include published context about human-derived samples or cell lines from the source studies.
+
+## Known issues
+
+- In `mem-cyclopropene-phosphine-conditions-ranking`, one `catalyst_scaffold` value in `protocol.json` joins the image path and a note (`structures/mem_catalyst.pngwithdrawnstereochemistry`). The image is `environment/public/structures/mem_catalyst.png`, and the experiment description links it correctly.
+- In `picalm-lipid-class-abundance`, the donor sex, age, ancestry, APOE genotype and cell-bank IDs given in the paper's runs were removed from `cell_context` in `protocol.json` and `experiment.md`. Donor codes, conditions, measurements and grading are unchanged.
+
+All other task files are as they were run in the paper.
+
+## Authors
+
+**Project Team:** Zengji Tu<sup>1*</sup>, Xiang Liu<sup>1*</sup>, Jiarui Sun<sup>1</sup>, Ruoyu Wu<sup>1</sup>, Zhihui Qi<sup>1</sup>, Chengda Wu<sup>1</sup>, Yutao Zhou<sup>1</sup>, Bo Yang<sup>1</sup>, Ziyan Zhang<sup>1</sup>, Jialong Wu<sup>1</sup>, Liang Chen<sup>1†‡</sup>, Kuan Li<sup>1†</sup>
+
+**Advisory Committee:** Steven Dillmann<sup>2</sup>, Yuanqi Du<sup>3</sup>, Soojung Yang<sup>4</sup>, Shizhe Diao<sup>5</sup>, Jianjun Wu<sup>6</sup>, Junlang Liu<sup>7</sup>, Xuming Hu<sup>8</sup>, Lin Chang<sup>6,9†</sup>, Baobao Chang<sup>10</sup>, Xiaowen Chu<sup>8</sup>, Philip S. Yu<sup>11</sup>, Tong-Yi Zhang<sup>8,12</sup>
+
+**Wet-Lab Collaborators:** Hao Cheng<sup>13</sup>, Qin Han<sup>13</sup>, Yanye Feng<sup>14</sup>, Chenru Duan<sup>15</sup>, Chuncheng Liu<sup>15</sup>, Yan Liu<sup>15</sup>, Yuchen Gao<sup>15</sup>, Zhao Li<sup>15</sup>, Ronghua Zhang<sup>15</sup>, Tory Xu<sup>15</sup>, Renjian Song<sup>16</sup>, Yaokai Fu<sup>16</sup>
+
+<sup>*</sup>Core contributors. <sup>†</sup>Corresponding authors: Liang Chen (<a href="mailto:liangchen@unipat.ai">liangchen@unipat.ai</a>), Kuan Li (<a href="mailto:kuanli@unipat.ai">kuanli@unipat.ai</a>), Lin Chang. <sup>‡</sup>Project leader: Liang Chen.
+
+<details>
+<summary>Affiliations</summary>
+
+1. UniPat AI
+2. Stanford University
+3. Cornell University
+4. FutureHouse
+5. The Hong Kong University of Science and Technology
+6. State Key Laboratory of Photonics and Communications, School of Electronics, Peking University
+7. Department of Chemistry and Chemical Biology, Harvard University
+8. The Hong Kong University of Science and Technology (Guangzhou)
+9. Frontiers Science Center for Nano-optoelectronics, Peking University
+10. Peking University
+11. University of Illinois Chicago
+12. Shanghai University
+13. Yurogen Biosystems LLC
+14. ABclonal Biotechnology Co., Ltd.
+15. Deep Principle
+16. Bota Biosciences
+
+</details>
+
+## Citation
+
+```bibtex
+@misc{spu2026,
+  title  = {Science Prediction Universe},
+  author = {Tu, Zengji and Liu, Xiang and Sun, Jiarui and Wu, Ruoyu and Qi, Zhihui and Wu, Chengda and Zhou, Yutao and Yang, Bo and Zhang, Ziyan and Wu, Jialong and Chen, Liang and Li, Kuan and Dillmann, Steven and Du, Yuanqi and Yang, Soojung and Diao, Shizhe and Wu, Jianjun and Liu, Junlang and Hu, Xuming and Chang, Lin and Chang, Baobao and Chu, Xiaowen and Yu, Philip S. and Zhang, Tong-Yi and Cheng, Hao and Han, Qin and Feng, Yanye and Duan, Chenru and Liu, Chuncheng and Liu, Yan and Gao, Yuchen and Li, Zhao and Zhang, Ronghua and Xu, Tory and Song, Renjian and Fu, Yaokai},
+  year   = {2026}
+}
+```

SOURCES.json

@@ -0,0 +1,2133 @@
+{
+ "tasks": [
+  {
+   "task_id": "amine-borane-arylation-bases",
+   "doi": "10.1038/s41586-025-09725-1",
+   "title": "Deaminative cross-coupling of amines by boryl radical β-scission",
+   "authors": [
+    "Zhenhua Zhang",
+    "Giovanni Lonardi",
+    "Thomas Sephton",
+    "Yusuf C. Guersoy",
+    "Chiara Stavagna",
+    "Giovanni V. A. Lenardon",
+    "Massimo Bietti",
+    "Daniele Leonori"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09725-1"
+  },
+  {
+   "task_id": "amine-borane-divergent-functionalization-ranking",
+   "doi": "10.1038/s41586-025-09725-1",
+   "title": "Deaminative cross-coupling of amines by boryl radical β-scission",
+   "authors": [
+    "Zhenhua Zhang",
+    "Giovanni Lonardi",
+    "Thomas Sephton",
+    "Yusuf C. Guersoy",
+    "Chiara Stavagna",
+    "Giovanni V. A. Lenardon",
+    "Massimo Bietti",
+    "Daniele Leonori"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09725-1"
+  },
+  {
+   "task_id": "bora-butenolide-conjugated-substrates",
+   "doi": "10.1038/s41467-025-60052-5",
+   "title": "Expedient access to bora-butenolide bioisosteres by counteranion-mediated trans-hydroboration of alkynes",
+   "authors": [
+    "Yuan-Wen Liu",
+    "Yu Liu",
+    "Yanting Zheng",
+    "Mengfan Zhang",
+    "Meng-En Ren",
+    "Peiyu Hua",
+    "Jie Han",
+    "Alois Fürstner",
+    "Hongming Jin"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-60052-5"
+  },
+  {
+   "task_id": "dialuminum-quinoline-catalyst-loading-yield",
+   "doi": "10.1038/s41467-025-63460-9",
+   "title": "Regio- and chemoselective catalytic partial transfer hydrogenation of quinolines by dinuclear aluminum species",
+   "authors": [
+    "Xufang Liu",
+    "Arseni Kostenko",
+    "Matthew M. D. Roy",
+    "Tobias Weng",
+    "Shigeyoshi Inoue"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-63460-9"
+  },
+  {
+   "task_id": "benzothiophene-photodimer-solvent-yield",
+   "doi": "10.1038/s41467-026-75078-6",
+   "title": "Visible Light-Induced Intermolecular Dearomatization for Constructing Angularly Fused Tetracyclic Scaffolds",
+   "authors": [
+    "Fan Lin",
+    "Ting-Ting Song",
+    "Li-Ming Zhang",
+    "Zhi-Yuan Ding",
+    "Yilitabaier Julaiti",
+    "Yanhua Lu",
+    "Boshun Wan",
+    "Yaguang Sun",
+    "Qing-An Chen"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-026-75078-6"
+  },
+  {
+   "task_id": "copper-allylic-amination-additive-yield",
+   "doi": "10.1038/s41467-025-64951-5",
+   "title": "Light-driven radical copper-catalyzed allylic amination via allylic copper intermediates",
+   "authors": [
+    "Hang Luo",
+    "Yupeng Yang",
+    "Yunpeng Ma",
+    "Fangnian Yu",
+    "Min Gao",
+    "Shizhao Xu",
+    "Huawei Ju",
+    "Yang Li",
+    "Kaifeng Wu",
+    "Luqing Lin"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-64951-5"
+  },
+  {
+   "task_id": "arene-acid-coupling-loading",
+   "doi": "10.1038/s41586-025-08887-2",
+   "title": "Generalizing arene C–H alkylations by radical–radical cross-coupling",
+   "authors": [
+    "Johannes Großkopf",
+    "Chawanansaya Gopatta",
+    "Robert T. Martin",
+    "Alexander Haseloer",
+    "David W. C. MacMillan"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-08887-2"
+  },
+  {
+   "task_id": "pyrazole-hydroamination-component-ranking",
+   "doi": "10.1038/s41586-026-10263-7",
+   "title": "Markovnikov hydroamination of terminal alkenes by phosphine redox catalysis",
+   "authors": [
+    "Flora Fan",
+    "Kassandra F. Sedillo",
+    "Alexander J. Maertens",
+    "Abigail G. Doyle"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-026-10263-7"
+  },
+  {
+   "task_id": "arene-acid-additional-catalysts-ranking",
+   "doi": "10.1038/s41586-025-08887-2",
+   "title": "Generalizing arene C–H alkylations by radical–radical cross-coupling",
+   "authors": [
+    "Johannes Großkopf",
+    "Chawanansaya Gopatta",
+    "Robert T. Martin",
+    "Alexander Haseloer",
+    "David W. C. MacMillan"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-08887-2"
+  },
+  {
+   "task_id": "rna-aminoacylation-strand-architecture-ranking",
+   "doi": "10.1038/s41586-025-09388-y",
+   "title": "Thioester-mediated RNA aminoacylation and peptidyl-RNA synthesis in water",
+   "authors": [
+    "Jyoti Singh",
+    "Benjamin Thoma",
+    "Daniel Whitaker",
+    "Max Satterly Webley",
+    "Yuan Yao",
+    "Matthew W. Powner"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09388-y"
+  },
+  {
+   "task_id": "rna-aminoacylation-donor-template-ranking",
+   "doi": "10.1038/s41586-025-09388-y",
+   "title": "Thioester-mediated RNA aminoacylation and peptidyl-RNA synthesis in water",
+   "authors": [
+    "Jyoti Singh",
+    "Benjamin Thoma",
+    "Daniel Whitaker",
+    "Max Satterly Webley",
+    "Yuan Yao",
+    "Matthew W. Powner"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09388-y"
+  },
+  {
+   "task_id": "alkene-transfer-conditions-ranking",
+   "doi": "10.1038/s41586-025-09159-9",
+   "title": "One-carbon homologation of alkenes",
+   "authors": [
+    "Marcus C. Grocott",
+    "Matthew J. Gaunt"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09159-9"
+  },
+  {
+   "task_id": "fluorene-dehydration-concentrations",
+   "doi": "10.1038/s41586-025-09490-1",
+   "title": "Robot-assisted mapping of chemical reaction hyperspaces and networks",
+   "authors": [
+    "Yankai Jia",
+    "Rafał Frydrych",
+    "Yaroslav I. Sobolev",
+    "Wai-Shing Wong",
+    "Bibek Prajapati",
+    "Daniel Matuszczyk",
+    "Yasemin Bilgi",
+    "Louis Gadina",
+    "Juan Carlos Ahumada",
+    "Galymzhan Moldagulov",
+    "Namhun Kim",
+    "Eric S. Larsen",
+    "Maxence Deschamps",
+    "Yanqiu Jiang",
+    "Bartosz A. Grzybowski"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09490-1"
+  },
+  {
+   "task_id": "azidopyridine-hydrolysis-condition-yield",
+   "doi": "10.1038/s41467-025-65313-x",
+   "title": "Formal pyridine meta-azidation and its application for the synthesis of diazepines, ring-fused δ-carbolines and 1,2,3-triazolylpyridines",
+   "authors": [
+    "Shu-Min Guo",
+    "Pengwei Xu",
+    "Constantin G. Daniliuc",
+    "Armido Studer"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-65313-x"
+  },
+  {
+   "task_id": "deutero-amination-reagent-solvent-yield",
+   "doi": "10.1038/s41467-025-57098-w",
+   "title": "Access to N-α-deuterated amino acids and DNA conjugates via Ca(II)-HFIP-mediated reductive deutero-amination of α-oxo-carbonyl compounds",
+   "authors": [
+    "Haoran Li",
+    "Yuwei Liu",
+    "Silin Zhang",
+    "Lei Ma",
+    "Zhongyi Zeng",
+    "Zhi Zhou",
+    "Vincent Gandon",
+    "Hongtao Xu",
+    "Wei Yi",
+    "Shengdong Wang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-57098-w"
+  },
+  {
+   "task_id": "indolylmethanol-cycloaddition-volume-temperature-yield",
+   "doi": "10.1038/s41467-025-62035-y",
+   "title": "Atroposelective construction of axially chiral alkenylindole-fused nine-membered rings via catalytic asymmetric formal (4 + 5) cycloaddition",
+   "authors": [
+    "Si-Jia Liu",
+    "Xue Wang",
+    "Ji-Xiang Yang",
+    "Xian-Song Ao",
+    "Shao-Fei Ni",
+    "Yu-Chen Zhang",
+    "Feng Shi"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-62035-y"
+  },
+  {
+   "task_id": "mem-cyclopropene-phosphine-conditions-ranking",
+   "doi": "10.1038/s41586-025-09485-y",
+   "title": "Catalytic enantioselective synthesis of alkylidenecyclopropanes",
+   "authors": [
+    "Jonathan C. Golec",
+    "Dong-Hang Tan",
+    "Ken Yamazaki",
+    "Eveline H. Tiekink",
+    "Kirsten E. Christensen",
+    "Trevor A. Hamlin",
+    "Darren J. Dixon"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09485-y"
+  },
+  {
+   "task_id": "bcb-aldehyde-substrate-isolated-yield-ranking",
+   "doi": "10.1038/s41467-025-64399-7",
+   "title": "Organocatalytic enantioselective [2π + 2σ] cycloaddition reactions of bicyclo[1.1.0]butanes with α,β-unsaturated aldehydes",
+   "authors": [
+    "Yi-Xiang Geng",
+    "Teng-Fei Xiao",
+    "Dong Xie",
+    "Ming-Ming Li",
+    "Pan-Pan Zhou",
+    "Guo-Qiang Xu",
+    "Peng-Fei Xu"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-64399-7"
+  },
+  {
+   "task_id": "nitronate-addition-preparative-scope-ranking",
+   "doi": "10.1038/s41586-025-09905-z",
+   "title": "The asymmetric synthesis of an acyclic N-stereogenic amine",
+   "authors": [
+    "Chendan Zhu",
+    "Sayantani Das",
+    "Marie Sophie Sterling",
+    "Nobuya Tsuji",
+    "Spencer J. Léger",
+    "Frank Neese",
+    "Chandra Kanta De",
+    "Benjamin List"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09905-z"
+  },
+  {
+   "task_id": "oxetane-catalyst-pair-yields",
+   "doi": "10.1038/s41586-025-09813-2",
+   "title": "Accelerating the discovery of multicatalytic cooperativity",
+   "authors": [
+    "Marcus H. Sak",
+    "Richard Y. Liu",
+    "Eugene E. Kwan",
+    "Eric N. Jacobsen"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09813-2"
+  },
+  {
+   "task_id": "nitronate-addition-additional-catalyst-ranking",
+   "doi": "10.1038/s41586-025-09905-z",
+   "title": "The asymmetric synthesis of an acyclic N-stereogenic amine",
+   "authors": [
+    "Chendan Zhu",
+    "Sayantani Das",
+    "Marie Sophie Sterling",
+    "Nobuya Tsuji",
+    "Spencer J. Léger",
+    "Frank Neese",
+    "Chandra Kanta De",
+    "Benjamin List"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09905-z"
+  },
+  {
+   "task_id": "organoboron-absolute-fluorescence-yield-ranking",
+   "doi": "10.1038/s41467-025-64698-z",
+   "title": "Rhodium-catalyzed construction of boron-based point and axial chirality via asymmetric annulation of alkynylborons",
+   "authors": [
+    "Xiaohan Zhu",
+    "Yue Shi",
+    "Jing Li",
+    "Haonan Peng",
+    "Mengxia Gong",
+    "Wei Zhou",
+    "Fen Wang",
+    "Genping Huang",
+    "Xingwei Li",
+    "Yu Fang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-64698-z"
+  },
+  {
+   "task_id": "fmoc-lysine-carbon-site-shift-ranking",
+   "doi": "10.1038/s41557-026-02112-x",
+   "title": "KDM3A catalyses the oxidation of acetyl-lysine to hydroxyacetyl-lysine on histone H3K9",
+   "authors": [
+    "Roman Belle",
+    "John-Paul Bukowski",
+    "Rachel Schiller",
+    "Ronald Cutler",
+    "Eidarus Salah",
+    "Robert S. Dawber",
+    "Anthony Tumber",
+    "Joanna Bonnici",
+    "Jessica D. Kindrick",
+    "Loane Serrano",
+    "Patrick Rabe",
+    "Catrine Johansson",
+    "Marie-Hélène Ruchaud",
+    "Richard J. Hopkinson",
+    "William D. Figg, Sr",
+    "Paul E. Brennan",
+    "David R. Mole",
+    "Simone Sidoli",
+    "Akane Kawamura",
+    "Christopher J. Schofield"
+   ],
+   "journal": "Nature Chemistry",
+   "url": "https://www.nature.com/articles/s41557-026-02112-x"
+  },
+  {
+   "task_id": "pyrrolidine-aza-shono-components-ranking",
+   "doi": "10.1038/s41586-024-08373-1",
+   "title": "Light-harvesting microelectronic devices for wireless electrosynthesis",
+   "authors": [
+    "Bartosz Górski",
+    "Jonas Rein",
+    "Samantha Norris",
+    "Yanxin Ji",
+    "Paul L. McEuen",
+    "Song Lin"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-024-08373-1"
+  },
+  {
+   "task_id": "benzoxazole-electroamination-library-ranking",
+   "doi": "10.1038/s41586-024-08373-1",
+   "title": "Light-harvesting microelectronic devices for wireless electrosynthesis",
+   "authors": [
+    "Bartosz Górski",
+    "Jonas Rein",
+    "Samantha Norris",
+    "Yanxin Ji",
+    "Paul L. McEuen",
+    "Song Lin"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-024-08373-1"
+  },
+  {
+   "task_id": "rhodium-alkenylphenol-alkene-substrate-isolated-yield-ranking",
+   "doi": "10.1038/s41467-025-59405-x",
+   "title": "Cathodic oxygen reduction-enabled rhodium-catalyzed (5 + 1) C–H/O–H annulation inspired by fuel cells",
+   "authors": [
+    "Yuan-Qiong Huang",
+    "Li Zhu",
+    "Tian-Sheng Mei"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-59405-x"
+  },
+  {
+   "task_id": "copper-cyanoesterification-condition-conversion-ranking",
+   "doi": "10.1038/s41467-025-62137-7",
+   "title": "Asymmetric cyanoesterification of vinylarenes by electrochemical copper catalysis",
+   "authors": [
+    "Kehan Zhou",
+    "Niankai Fu"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-62137-7"
+  },
+  {
+   "task_id": "dabconium-oxidant-electrode-nmr-yield-ranking",
+   "doi": "10.1038/s44160-025-00890-9",
+   "title": "Electrochemical oxidation enables aromatic C–H amination with dual mechanisms",
+   "authors": [
+    "Griffin Stewart",
+    "Eva Maria Alvarez",
+    "Chris Rapala",
+    "Jonathan H. Sklar",
+    "Julia A. Kalow",
+    "Christian A. Malapit"
+   ],
+   "journal": "Nature Synthesis",
+   "url": "https://www.nature.com/articles/s44160-025-00890-9"
+  },
+  {
+   "task_id": "ionic-cof-directional-solvent-permeance-ranking",
+   "doi": "10.1038/s41467-025-63240-5",
+   "title": "Asymmetric ionic covalent organic framework membranes with tunable Turing patterns prepared via ion regulated strategy",
+   "authors": [
+    "Jingcheng Du",
+    "Jian guan",
+    "Ali A. AL-Thuraya",
+    "Weiwang Lim",
+    "Qian Sun",
+    "Ayan Yao",
+    "Dong Cao",
+    "Ji Ma",
+    "Yuting zhang",
+    "Yumo Fan",
+    "Pengjia Dou",
+    "Jiangtao Liu"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-63240-5"
+  },
+  {
+   "task_id": "silver-ironoxyhydroxide-propylene-epoxidation-ranking",
+   "doi": "10.1038/s41467-025-67338-8",
+   "title": "In-situ spectroscopic insights into the dual-site synergistic electrocatalytic mechanism of propylene epoxidation over single-Ag-atom catalyst",
+   "authors": [
+    "Dengwen Pi",
+    "Xiaobo Yang",
+    "Meixin Chen",
+    "Jian Zhao",
+    "Jiuyi Wang",
+    "Shifu Wang",
+    "Wen-Cheng Shih",
+    "Wei Xu",
+    "Yanqiang Huang",
+    "Bin Liu",
+    "Xuning Li"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-67338-8"
+  },
+  {
+   "task_id": "palladium-bromoalkene-substrate-cyclization-isolated-yield-ranking",
+   "doi": "10.1038/s41467-025-58633-5",
+   "title": "Synthesis of benzoheterocycles by palladium-catalyzed migratory cyclization through an unexpected reaction cascade",
+   "authors": [
+    "Wen-Cong Li",
+    "Lin Zhang",
+    "Shiming Bai",
+    "Jia-Hao Zhao",
+    "Guang-Rui Liu",
+    "Yu Lan",
+    "Shufeng Chen",
+    "Jialin Ming"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-58633-5"
+  },
+  {
+   "task_id": "alkyl-iodide-nickel-difluoromethylation-additive-yield",
+   "doi": "10.1038/s41929-026-01505-y",
+   "title": "Enantioselective difluoromethylation of unactivated alkyl halides via a formal nickel(II/IV) cycle",
+   "authors": [
+    "Xian Zhao",
+    "Lingfeng Yin",
+    "Chia-Jung Yang",
+    "Andrew T. Poore",
+    "Xinyu Zhang",
+    "Stephen C. Yachuw",
+    "Chao Wang",
+    "Xiao Wang",
+    "Hairong Guan",
+    "Jeanette A. Krause",
+    "Mu-Jeng Cheng",
+    "Shiliang Tian",
+    "Wei Liu"
+   ],
+   "journal": "Nature Catalysis",
+   "url": "https://www.nature.com/articles/s41929-026-01505-y"
+  },
+  {
+   "task_id": "alkyne-bridged-ester-palladium-screen",
+   "doi": "10.1038/s41467-025-64279-0",
+   "title": "Modular esterification of unstrained carbonyls through palladium-catalyzed alkyne bridging C-C bond activation",
+   "authors": [
+    "Mingruo Ding",
+    "Mengmeng Niu",
+    "Lingfei Hu",
+    "Qifan Guan",
+    "Shan Yuan",
+    "Yinghua Yu",
+    "Maojun Xiao",
+    "Shenlin Huang",
+    "Gang Lu",
+    "Xueliang Huang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-64279-0"
+  },
+  {
+   "task_id": "bromovinyl-phosphine-migratory-cyclization-yield",
+   "doi": "10.1038/s41467-025-58633-5",
+   "title": "Synthesis of benzoheterocycles by palladium-catalyzed migratory cyclization through an unexpected reaction cascade",
+   "authors": [
+    "Wen-Cong Li",
+    "Lin Zhang",
+    "Shiming Bai",
+    "Jia-Hao Zhao",
+    "Guang-Rui Liu",
+    "Yu Lan",
+    "Shufeng Chen",
+    "Jialin Ming"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-58633-5"
+  },
+  {
+   "task_id": "pd-sulfoximine-solvent-cyclization-yield",
+   "doi": "10.1038/s41467-025-57251-5",
+   "title": "Asymmetric S=N-embedded polyaromatic construction via enantioselective Pd-catalyzed C–H activation",
+   "authors": [
+    "Daming Zeng",
+    "Xinyu Zhang",
+    "Ming Wang",
+    "Xuefeng Jiang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-57251-5"
+  },
+  {
+   "task_id": "cobalt-vinyl-acetate-ligand-conversion",
+   "doi": "10.1038/s41467-025-62145-7",
+   "title": "A SaBOX/Co catalyst designed for facile access to the challenging copolymers of vinyl ester and methacrylate",
+   "authors": [
+    "Yi-Jie Ding",
+    "Sheng-Ye Zhang",
+    "Ya-Ning Li",
+    "Shu-Yang Yu",
+    "Kongchanghao Huang",
+    "Yuchen Zhang",
+    "Jun-Fang Li",
+    "Xiao-Yan Wang",
+    "Yong Tang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-62145-7"
+  },
+  {
+   "task_id": "ethylene-ctfe-solvent-polymer-mass",
+   "doi": "10.1038/s41467-025-66962-8",
+   "title": "Photoredox-controlled alternating copolymerization enables highly crystalline structures and block copolymers from thermoplastic to elastomer",
+   "authors": [
+    "Mengli Xu",
+    "Qiankai Chen",
+    "Shantao Han",
+    "Kaixuan Chen",
+    "Xing Guo",
+    "Mao Chen"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-66962-8"
+  },
+  {
+   "task_id": "cobalt-methyl-methacrylate-catalyst-molar-mass-ranking",
+   "doi": "10.1038/s41467-025-62145-7",
+   "title": "A SaBOX/Co catalyst designed for facile access to the challenging copolymers of vinyl ester and methacrylate",
+   "authors": [
+    "Yi-Jie Ding",
+    "Sheng-Ye Zhang",
+    "Ya-Ning Li",
+    "Shu-Yang Yu",
+    "Kongchanghao Huang",
+    "Yuchen Zhang",
+    "Jun-Fang Li",
+    "Xiao-Yan Wang",
+    "Yong Tang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-62145-7"
+  },
+  {
+   "task_id": "polymer-diels-alder-concentration-dose-yield-ranking",
+   "doi": "10.1038/s41467-025-65707-x",
+   "title": "Essential oil-derived decomposable polymers via cycloaddition polymerization of silyl ether-linked phenylpropanoids",
+   "authors": [
+    "Ryo Nagaya",
+    "Tatsuya Seko",
+    "Kazuhiro Okamoto",
+    "Kazuhide Ueno",
+    "Mahito Atobe",
+    "Naoki Shida"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-65707-x"
+  },
+  {
+   "task_id": "oxychloride-precursor-milling-conductivity-ranking",
+   "doi": "10.1038/s41467-025-68210-5",
+   "title": "Mechanically compliant and cost-effective 1.4Li2O-0.75ZrCl4-0.25AlCl3 solid electrolyte for all-solid-state batteries with improved cycling stability",
+   "authors": [
+    "Lv Hu",
+    "Yaolong He",
+    "Dong Wang",
+    "Wanxia Li",
+    "Jingming Yao",
+    "Xiaolong Zhang",
+    "Jinfeng Zhu",
+    "Huaican Chen",
+    "Wen Yin",
+    "Yanru Wang",
+    "Kejun Yan",
+    "Jinzhu Wang",
+    "Hui Li",
+    "Fang Chen",
+    "Yating Liu",
+    "Junqi Lai",
+    "Qi Chen",
+    "Jie Ma",
+    "Shuhong Jiao",
+    "Guorui Wang",
+    "Siqi Shi",
+    "Liwei Chen",
+    "Jianyu Huang",
+    "Cheng Ma"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-68210-5"
+  },
+  {
+   "task_id": "zinc-additive-eqcm-full-trace-response-ranking",
+   "doi": "10.1038/s41467-026-70366-7",
+   "title": "Electric-field-reinforced affinitive electrolytes for highly reversible aqueous zinc metal batteries",
+   "authors": [
+    "Ming Yang",
+    "Bao Zhang",
+    "An Duan",
+    "Sha Luo",
+    "Luyang Ge",
+    "Jiaqi Wang",
+    "Yuxi Zhang",
+    "Yu Feng",
+    "Xiao Huang",
+    "Yuyang Tang",
+    "Jia Yao",
+    "Hao Wang",
+    "Houzhao Wan",
+    "Wei Sun"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-026-70366-7"
+  },
+  {
+   "task_id": "zinc-electrolyte-organic-additive-contact-angle-ranking",
+   "doi": "10.1038/s41565-026-02187-0",
+   "title": "Hydrophobic liquid electrolyte interphases for efficient aqueous zinc batteries",
+   "authors": [
+    "Guanjie Li",
+    "Shilin Zhang",
+    "Jodie A. Yuwono",
+    "Xinyu Li",
+    "Javen Qinfeng Shi",
+    "Chunsheng Wang",
+    "Zaiping Guo"
+   ],
+   "journal": "Nature Nanotechnology",
+   "url": "https://www.nature.com/articles/s41565-026-02187-0"
+  },
+  {
+   "task_id": "ether-zinc-electrolyte-concentration-lifespan-ranking",
+   "doi": "10.1038/s41565-026-02187-0",
+   "title": "Hydrophobic liquid electrolyte interphases for efficient aqueous zinc batteries",
+   "authors": [
+    "Guanjie Li",
+    "Shilin Zhang",
+    "Jodie A. Yuwono",
+    "Xinyu Li",
+    "Javen Qinfeng Shi",
+    "Chunsheng Wang",
+    "Zaiping Guo"
+   ],
+   "journal": "Nature Nanotechnology",
+   "url": "https://www.nature.com/articles/s41565-026-02187-0"
+  },
+  {
+   "task_id": "metal-dicarboxylate-mof-first-capacity",
+   "doi": "10.1038/s41467-025-64711-5",
+   "title": "Low-strain metal–organic framework negative electrode for stable all-solid-state batteries",
+   "authors": [
+    "Minje Ryu",
+    "Sung-O Park",
+    "Seongje Lim",
+    "Hwasoo Lee",
+    "Jung Hwan Lee",
+    "Gwanghyun Lee",
+    "Ki Yoon Bae",
+    "Ji Young Kim",
+    "Youngsun Kong",
+    "Samick Son",
+    "Kisuk Kang",
+    "Jong Hyeok Park"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-64711-5"
+  },
+  {
+   "task_id": "metal-oxide-iridium-oxygen-evolution-current-ranking",
+   "doi": "10.1038/s41565-026-02136-x",
+   "title": "Proton–electron temporal asynchrony on femtosecond timescales enables anti-corrosive low-iridium anodes for PEM electrolysers",
+   "authors": [
+    "Wei Shen",
+    "Fei-Yue Gao",
+    "Xiaogang Sun",
+    "Haodian Xie",
+    "Yang Hu",
+    "Huiying Wu",
+    "Mietek Jaroniec",
+    "Yao Zheng",
+    "Pinxian Xi",
+    "Chun-Hua Yan",
+    "Shi-Zhang Qiao"
+   ],
+   "journal": "Nature Nanotechnology",
+   "url": "https://www.nature.com/articles/s41565-026-02136-x"
+  },
+  {
+   "task_id": "silver-nickel-gasfed-electrolyte-co-efficiency-ranking",
+   "doi": "10.1038/s41467-025-65331-9",
+   "title": "Efficient amino-acid-based reactive capture of CO2 via nickel molecular catalyst",
+   "authors": [
+    "Zunmin Guo",
+    "Feng Li",
+    "Yurou Celine Xiao",
+    "Sung-Fu Hung",
+    "Ying-Rui Lu",
+    "Amir Foroozan",
+    "Jieyuan Liu",
+    "Siyu Sonia Sun",
+    "Shijie Liu",
+    "Yuxuan Che",
+    "Qiyou Wang",
+    "Min Liu",
+    "Cai Wang",
+    "Yuke Li",
+    "Kang-Shun Peng",
+    "Yu-Cheng Liu",
+    "Mengyang Fan",
+    "Zahra Azimi Dijvejin",
+    "Panagiotis Papangelakis",
+    "Yong Wang",
+    "Ali Shayesteh Zeraati",
+    "Kai Han",
+    "Paul Corbett",
+    "Drew Higgins",
+    "Rui Kai Miao",
+    "David Sinton"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-65331-9"
+  },
+  {
+   "task_id": "metal-zno-methane-product-ranking",
+   "doi": "10.1038/s41467-025-57854-y",
+   "title": "Photothermal direct methane conversion to formaldehyde at the gas-solid interface under ambient pressure",
+   "authors": [
+    "Yuxiong Wang",
+    "Yaoyu Zhang",
+    "Xiaoqiang Wang",
+    "Yue Liu",
+    "Zhongbiao Wu"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-57854-y"
+  },
+  {
+   "task_id": "mos2-znis-loading-hydrogen-180min",
+   "doi": "10.1038/s41467-025-66003-4",
+   "title": "Dipole synergy enables fast directional charge transport for solar hydrogen and benzaldehyde coproduction",
+   "authors": [
+    "Zhennan Wang",
+    "Dingyanyan Zhou",
+    "Kaige Tian",
+    "Guilin Chen",
+    "Youyong Li",
+    "Shengzhong Frank Liu",
+    "Shuit-Tong Lee",
+    "Yujin Ji",
+    "Junqing Yan"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-66003-4"
+  },
+  {
+   "task_id": "cuau-isotope-hydrogen-incorporation-ranking",
+   "doi": "10.1038/s41467-025-67317-z",
+   "title": "Catalyst alloying enables control over the electrochemical hydrogen transfer route",
+   "authors": [
+    "Jiguang Zhang",
+    "Chengyi Zhang",
+    "Weng Weei Tjiu",
+    "Tingting Fan",
+    "Yu-Chia Chang",
+    "Yu-Ting Chueh",
+    "Sibo Wang",
+    "Yan Zhang",
+    "Meng Wang",
+    "Qin Yang",
+    "Zeping Wang",
+    "Ruoou Yang",
+    "Zainul Aabdin",
+    "Sung-Fu Hung",
+    "Ziyun Wang",
+    "Yanwei Lum"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-67317-z"
+  },
+  {
+   "task_id": "perovskite-ligand-carrier-lifetime",
+   "doi": "10.1038/s41586-026-10626-0",
+   "title": "Stereoelectronic manipulation of ligands for perovskite solar cells",
+   "authors": [
+    "Tinghuan Yang",
+    "Erxin Zhao",
+    "Nan Wu",
+    "Xiaoming Chang",
+    "Chenqing Tian",
+    "Hai-Long Wang",
+    "Lu Zhang",
+    "Nannan Gu",
+    "Ting Nie",
+    "Ye Yang",
+    "Zheng Zhang",
+    "Tianfei Xu",
+    "Xin Chen",
+    "Shuang Wang",
+    "Tianqi Niu",
+    "Niansheng Xu",
+    "Chuang Ma",
+    "Haojin Li",
+    "Buyi Yan",
+    "Zicheng Ding",
+    "Shengzhong Frank Liu",
+    "Feng Gao",
+    "Kui Zhao"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-026-10626-0"
+  },
+  {
+   "task_id": "hybrid-sam-coadsorbent-photovoltaic-ranking",
+   "doi": "10.1038/s41467-025-58111-y",
+   "title": "Modulating competitive adsorption of hybrid self-assembled molecules for efficient wide-bandgap perovskite solar cells and tandems",
+   "authors": [
+    "Chenyang Shi",
+    "Jianan Wang",
+    "Xia Lei",
+    "Qisen Zhou",
+    "Weitao Wang",
+    "Zhichun Yang",
+    "Sanwan Liu",
+    "Jiaqi Zhang",
+    "He Zhu",
+    "Rui Chen",
+    "Yongyan Pan",
+    "Zhengtian Tan",
+    "Wenguang Liu",
+    "Zhengjing Zhao",
+    "Zihe Cai",
+    "Xiaojun Qin",
+    "Zhiguo Zhao",
+    "Jingbai Li",
+    "Zonghao Liu",
+    "Wei Chen"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-58111-y"
+  },
+  {
+   "task_id": "amino-acid-peptide-nanotube-emission-band-ratio-ranking",
+   "doi": "10.1038/s44160-026-01054-z",
+   "title": "Handle-free attachment of small molecules on single-walled carbon nanotubes",
+   "authors": [
+    "Stanislav S. Piletsky",
+    "Erin E. Keblish",
+    "Alec R. Goffin",
+    "Xiaojia Jin",
+    "In-Jun Hwang",
+    "August Amb",
+    "Dmitri Kilin",
+    "Svetlana Kilina",
+    "Ming Zheng",
+    "Mijin Kim",
+    "Daniel A. Heller"
+   ],
+   "journal": "Nature Synthesis",
+   "url": "https://www.nature.com/articles/s44160-026-01054-z"
+  },
+  {
+   "task_id": "pvdf-trfe-press-temperature-voltage",
+   "doi": "10.1038/s41467-025-67519-5",
+   "title": "Thermomechanical pressing and immediate quenching: enhanced piezoelectricity and transparency in piezoelectric nanofibers",
+   "authors": [
+    "Zhan Qu",
+    "Qian Chen",
+    "Xinyu Li",
+    "Jinyi Liu",
+    "Chenhong Lang",
+    "Yu Xu",
+    "Rujun Tang",
+    "Jizhen Zhang",
+    "Hongxia Wang",
+    "Yong Du",
+    "Tong Lin",
+    "Jian Fang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-67519-5"
+  },
+  {
+   "task_id": "mownb-mica-compressive-temperature-response",
+   "doi": "10.1038/s41467-025-62100-6",
+   "title": "Highly customizable, ultrawide-temperature free-form flexible sensing electronic systems based on medium-entropy alloy paintings",
+   "authors": [
+    "Weiwei Li",
+    "Yingzhe Li",
+    "Manzhang Xu",
+    "Yilin Zhou",
+    "Ruoyan Miao",
+    "Kexin Wang",
+    "Yunqiang Cao",
+    "Yizhong Song",
+    "Siying Dang",
+    "Lu Zheng",
+    "Xuewen Wang",
+    "Wei Huang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-62100-6"
+  },
+  {
+   "task_id": "metal-zif-nitrogen-adsorption-desorption-ranking",
+   "doi": "10.1038/s41467-025-62110-4",
+   "title": "Multivariate metal-organic frameworks enable chemical shift-encoded MRI with femtomolar sensitivity for biological systems",
+   "authors": [
+    "Qingbin Zeng",
+    "Quer Yue",
+    "Ming Zhang",
+    "Rui Zhou",
+    "Zhen Wang",
+    "Zhongyu Xiao",
+    "Yaping Yuan",
+    "Xiuchao Zhao",
+    "Weiping Jiang",
+    "Lei Zhang",
+    "Yuqi Yang",
+    "Minghui Yang",
+    "Haidong Li",
+    "Louis-S. Bouchard",
+    "Qianni Guo",
+    "Xin Zhou"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-62110-4"
+  },
+  {
+   "task_id": "oil-filter-air-transport-ranking",
+   "doi": "10.1038/s41586-025-09156-y",
+   "title": "Bioinspired capillary force-driven super-adhesive filter",
+   "authors": [
+    "Junyong Park",
+    "Chan Sik Moon",
+    "Ji Min Lee",
+    "Sazzadul A. Rahat",
+    "Sang Moon Kim",
+    "Jonathan T. Pham",
+    "Michael Kappl",
+    "Hans-Jürgen Butt",
+    "Sanghyuk Wooh"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09156-y"
+  },
+  {
+   "task_id": "pse-resin-ion-resistance-ranking",
+   "doi": "10.1038/s41467-025-58385-2",
+   "title": "Optimization and scaling-up of porous solid electrolyte electrochemical reactors for hydrogen peroxide electrosynthesis",
+   "authors": [
+    "Erzhuo Zhao",
+    "Yixin Zhang",
+    "Juhong Zhan",
+    "Guangsen Xia",
+    "Gang Yu",
+    "Yujue Wang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-58385-2"
+  },
+  {
+   "task_id": "pva-cacl2-mineralization-mechanics",
+   "doi": "10.1038/s41467-025-66423-2",
+   "title": "Tough and tear resistant hydrogel with a sandwich mineralized structure induced by bidirectional ion migration",
+   "authors": [
+    "Dongsheng Yang",
+    "Zhiyu Zhao",
+    "Liangyu Zhu",
+    "Wenbin Tang",
+    "Zirou Qiu",
+    "Yifan Wu",
+    "Wenchuan Jiang",
+    "Cheng Lan",
+    "Yifu Zhang",
+    "Tao Liu",
+    "Heping Xie"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-66423-2"
+  },
+  {
+   "task_id": "hydrogel-additional-formulation-adhesion",
+   "doi": "10.1038/s41586-025-09269-4",
+   "title": "Data-driven de novo design of super-adhesive hydrogels",
+   "authors": [
+    "Hongguang Liao",
+    "Sheng Hu",
+    "Hu Yang",
+    "Lei Wang",
+    "Shinya Tanaka",
+    "Ichigaku Takigawa",
+    "Wei Li",
+    "Hailong Fan",
+    "Jian Ping Gong"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09269-4"
+  },
+  {
+   "task_id": "hydrogel-formulation-oscillatory-rheology",
+   "doi": "10.1038/s41586-025-09269-4",
+   "title": "Data-driven de novo design of super-adhesive hydrogels",
+   "authors": [
+    "Hongguang Liao",
+    "Sheng Hu",
+    "Hu Yang",
+    "Lei Wang",
+    "Shinya Tanaka",
+    "Ichigaku Takigawa",
+    "Wei Li",
+    "Hailong Fan",
+    "Jian Ping Gong"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09269-4"
+  },
+  {
+   "task_id": "tripeptide-dried-assembly-afm-modulus-ranking",
+   "doi": "10.1038/s41563-025-02300-z",
+   "title": "Adaptive peptide dispersions enable drying-induced biomolecule encapsulation",
+   "authors": [
+    "Dhwanit R. Dave",
+    "Salma Kassem",
+    "Maeva Coste",
+    "Lele Xu",
+    "Mona Tayarani-Najjaran",
+    "Darjan Podbevšek",
+    "Paola Colon-De Leon",
+    "Sheng Zhang",
+    "Luis Ortuno Macias",
+    "Deborah Sementa",
+    "María Pérez-Ferreiro",
+    "Nooshin Sadat Ayati",
+    "Muniyat A. Choudhury",
+    "Kelly Veerasammy",
+    "Selma Doganata",
+    "Tiffany Zhong",
+    "Cory Weng",
+    "Jorge Morales",
+    "Denize C. Favaro",
+    "Mateusz Marianski",
+    "So Yeon Ahn",
+    "Allie C. Obermeyer",
+    "Tong Wang",
+    "Tai-De Li",
+    "Xi Chen",
+    "Raymond Tu",
+    "Ye He",
+    "Rein V. Ulijn"
+   ],
+   "journal": "Nature Materials",
+   "url": "https://www.nature.com/articles/s41563-025-02300-z"
+  },
+  {
+   "task_id": "twisted-plywood-matrix-angle-fracture-ranking",
+   "doi": "10.1038/s41467-024-55723-8",
+   "title": "Distorting crack-front geometry for enhanced toughness by manipulating bioinspired heterogeneity",
+   "authors": [
+    "Kaijin Wu",
+    "Zhaoqiang Song",
+    "Mengqi Liu",
+    "Zewen Wang",
+    "Si-Ming Chen",
+    "Shu-Hong Yu",
+    "Linghui He",
+    "Yong Ni"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-024-55723-8"
+  },
+  {
+   "task_id": "lipoic-acid-dbu-molar-ratio-adhesion-ranking",
+   "doi": "10.1038/s41467-025-62917-1",
+   "title": "Reversible biobased adhesives enable closed-loop engineered composites",
+   "authors": [
+    "Jin Lv",
+    "Daxin Zhang",
+    "Xinkai Li",
+    "Yinggang Miao",
+    "Yuyan Wang",
+    "Ying Wang",
+    "Xinxing Zhang"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-62917-1"
+  },
+  {
+   "task_id": "underwater-adhesive-prototype-configuration-ranking",
+   "doi": "10.1038/s41586-025-09304-4",
+   "title": "Mechanical underwater adhesive devices for soft substrates",
+   "authors": [
+    "Ziliang Kang",
+    "Johanna A. Gomez",
+    "Alisa MeiShan Ross",
+    "Ameya R. Kirtane",
+    "Ming Zhao",
+    "Yubin Cai",
+    "Fu Xing Chen",
+    "Corona L. Chen",
+    "Isaac Diaz Becdach",
+    "Rajib Dey",
+    "Andrei Russel Ismael",
+    "Injoo Moon",
+    "Yiyuan Yang",
+    "Benjamin N. Muller",
+    "Mehmet Girayhan Say",
+    "Andrew Pettinari",
+    "Jason Kobrin",
+    "Joshua Morimoto",
+    "Ted Smierciak",
+    "Aaron Lopes",
+    "Ayten Ebru Erdogan",
+    "Matt Murphy",
+    "Niora Fabian",
+    "Ashley Guevara",
+    "Benedict Laidlaw",
+    "Kailyn Schmidt",
+    "Alison M. Hayward",
+    "Alexandra H. Techet",
+    "Christopher P. Kenaley",
+    "Giovanni Traverso"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09304-4"
+  },
+  {
+   "task_id": "agcute-bending-endpoints-ranking",
+   "doi": "10.1038/s41467-025-58104-x",
+   "title": "Strategic vacancy engineering advances record-high ductile AgCu(Te, Se, S) thermoelectrics",
+   "authors": [
+    "Nan-Hai Li",
+    "Xiao-Lei Shi",
+    "Si-Qi Liu",
+    "Meng Li",
+    "Tian-Yi Cao",
+    "Min Zhang",
+    "Wan-Yu Lyu",
+    "Wei-Di Liu",
+    "Dong-Chen Qi",
+    "Zhi-Gang Chen"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-58104-x"
+  },
+  {
+   "task_id": "fenicoalta-processing-tensile-properties",
+   "doi": "10.1038/s41586-025-09160-2",
+   "title": "Machine-learning design of ductile FeNiCoAlTa alloys with high strength",
+   "authors": [
+    "Yasir Sohail",
+    "Chongle Zhang",
+    "Dezhen Xue",
+    "Jinyu Zhang",
+    "Dongdong Zhang",
+    "Shaohua Gao",
+    "Yang Yang",
+    "Xiaoxuan Fan",
+    "Hang Zhang",
+    "Gang Liu",
+    "Jun Sun",
+    "En Ma"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09160-2"
+  },
+  {
+   "task_id": "fenicoalta-tensile-compositions",
+   "doi": "10.1038/s41586-025-09160-2",
+   "title": "Machine-learning design of ductile FeNiCoAlTa alloys with high strength",
+   "authors": [
+    "Yasir Sohail",
+    "Chongle Zhang",
+    "Dezhen Xue",
+    "Jinyu Zhang",
+    "Dongdong Zhang",
+    "Shaohua Gao",
+    "Yang Yang",
+    "Xiaoxuan Fan",
+    "Hang Zhang",
+    "Gang Liu",
+    "Jun Sun",
+    "En Ma"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09160-2"
+  },
+  {
+   "task_id": "empty-lipid-nanoparticle-zeta-potential-ranking",
+   "doi": "10.1038/s41565-025-01958-5",
+   "title": "Tailoring the adjuvanticity of lipid nanoparticles by PEG lipid ratio and phospholipid modifications",
+   "authors": [
+    "Máté Vadovics",
+    "Wenchen Zhao",
+    "Emily F. Daley",
+    "Kieu Lam",
+    "Owen Daly",
+    "Khalid Rashid",
+    "Hailey R. Lee",
+    "Petra Schreiner",
+    "Kendall A. Lundgreen",
+    "Brian T. Gaudette",
+    "Vladimir V. Shuvaev",
+    "Evguenia Arguiri",
+    "Hiromi Muramatsu",
+    "András Sárközy",
+    "Thandiswa Mdluli",
+    "Junchao Xu",
+    "Xuexiang Han",
+    "Nina De Luna",
+    "Diana Castaño",
+    "Emily Bettini",
+    "Edit Ábrahám",
+    "Zoltan Lipinszki",
+    "Giuseppe Carlucci",
+    "Avinash Haridas Bansode",
+    "Katelyn Nguyen",
+    "Thuc M. Le",
+    "Tony Luu",
+    "Vladimir R. Muzykantov",
+    "Paul Bates",
+    "David Allman",
+    "Michael J. Mitchell",
+    "Michela Locci",
+    "Caius G. Radu",
+    "James Heyes",
+    "Norbert Pardi"
+   ],
+   "journal": "Nature Nanotechnology",
+   "url": "https://www.nature.com/articles/s41565-025-01958-5"
+  },
+  {
+   "task_id": "gastruloid-signalling-cell-count-ranking",
+   "doi": "10.1038/s41556-026-01923-x",
+   "title": "Mosaic gastruloids reveal a temporal restriction for developmental cell competition",
+   "authors": [
+    "Joshua D. Frenster",
+    "Stephen Babin",
+    "Pablo Casani-Galdon",
+    "Joel B. Josende-Garcia",
+    "Pau Pascual-Mas",
+    "Gaëlle Robertson",
+    "Shlomit Edri",
+    "Alexandra E. Wehmeyer",
+    "Sebastian J. Arnold",
+    "Jordi Garcia Ojalvo",
+    "Alfonso Martinez Arias"
+   ],
+   "journal": "Nature Cell Biology",
+   "url": "https://www.nature.com/articles/s41556-026-01923-x"
+  },
+  {
+   "task_id": "fmn-aptamer-fluorescence-ranking",
+   "doi": "10.1038/s41467-025-60722-4",
+   "title": "Rational development of FMN-based orthogonal riboswitch that functions in response to specific non-cognate ligand",
+   "authors": [
+    "Divya Ojha",
+    "Ambadas B. Rode"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-60722-4"
+  },
+  {
+   "task_id": "bik1-positional-peptide-library-phosphorylation-ranking",
+   "doi": "10.1038/s41477-025-02218-z",
+   "title": "Motif-based substrate mapping of the receptor-like cytoplasmic kinase BIK1 reveals novel components and regulatory nodes of plant immunity",
+   "authors": [
+    "Ryan Toth",
+    "Sera Choi",
+    "Marie Le Naour--Vernet",
+    "Florian Schwanke",
+    "Jared L. Johnson",
+    "Estee E. Tee",
+    "Tomer M. Yaron-Barir",
+    "Eleanor Khochaba",
+    "Paul Derbyshire",
+    "Anka Colo",
+    "Philipp Köster",
+    "Emily M. Huntsman",
+    "Laura Herold",
+    "Yoonyoung Lee",
+    "Álvaro D. Fernández-Fernández",
+    "Hee-Kyung Ahn",
+    "Julian Dindas",
+    "Marta Bjornson",
+    "Jack Rhodes",
+    "Beibei Song",
+    "Weibing Wang",
+    "Marija Smokvarska",
+    "Emmanuelle M. Bayer",
+    "Jian-Min Zhou",
+    "Lewis C. Cantley",
+    "Jonathan D. G. Jones",
+    "Kyle W. Bender",
+    "Frank L. H. Menke",
+    "Christine Faulkner",
+    "Cyril Zipfel",
+    "Thomas A. DeFalco"
+   ],
+   "journal": "Nature Plants",
+   "url": "https://www.nature.com/articles/s41477-025-02218-z"
+  },
+  {
+   "task_id": "tomato-multilocus-field-branching",
+   "doi": "10.1038/s41586-025-09243-0",
+   "title": "Cryptic variation fuels plant phenotypic change through hierarchical epistasis",
+   "authors": [
+    "Sophia G. Zebell",
+    "Carlos Martí-Gómez",
+    "Blaine Fitzgerald",
+    "Camila P. Cunha",
+    "Michael Lach",
+    "Brooke M. Seman",
+    "Anat Hendelman",
+    "Simon Sretenovic",
+    "Yiping Qi",
+    "Madelaine Bartlett",
+    "Yuval Eshed",
+    "David M. McCandlish",
+    "Zachary B. Lippman"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09243-0"
+  },
+  {
+   "task_id": "arabidopsis-centromere-recombination-ranking",
+   "doi": "10.1038/s41477-026-02329-1",
+   "title": "The deSUMOylase SPF2 and the cohesin regulators SGO2 and CTF18 suppress crossovers near centromeres",
+   "authors": [
+    "Rigel Salinas Gamboa",
+    "Joiselle Blanche Fernandes",
+    "Qichao Lian",
+    "Dipesh Kumar Singh",
+    "Stephanie Durand",
+    "Divyanshu Sahu",
+    "Laia Capilla-Perez",
+    "Hernan Lopez",
+    "Alexandra Kalde",
+    "Raphael Mercier"
+   ],
+   "journal": "Nature Plants",
+   "url": "https://www.nature.com/articles/s41477-026-02329-1"
+  },
+  {
+   "task_id": "arabidopsis-amino-acid-dark-chlorophyll-ranking",
+   "doi": "10.1038/s41477-026-02328-2",
+   "title": "The progression of leaf senescence is gated by the cytosolic arginine pool",
+   "authors": [
+    "Shah Hussain",
+    "Clément Boussardon",
+    "Olivier Keech"
+   ],
+   "journal": "Nature Plants",
+   "url": "https://www.nature.com/articles/s41477-026-02328-2"
+  },
+  {
+   "task_id": "sgc-mitochondrial-transfer-ranking",
+   "doi": "10.1038/s41586-025-09896-x",
+   "title": "Mitochondrial transfer from glia to neurons protects against peripheral neuropathy",
+   "authors": [
+    "Jing Xu",
+    "Yize Li",
+    "Charles Novak",
+    "Min Lee",
+    "Zihan Yan",
+    "Sangsu Bang",
+    "Aidan McGinnis",
+    "Sharat Chandra",
+    "Vivian Zhang",
+    "Wei He",
+    "Terry Lechler",
+    "Maria Pia Rodriguez Salazar",
+    "Cagla Eroglu",
+    "Matthew L. Becker",
+    "Dmitry Velmeshev",
+    "Richard E. Cheney",
+    "Ru-Rong Ji"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09896-x"
+  },
+  {
+   "task_id": "peptide-fos-expression-ranking",
+   "doi": "10.1038/s41586-025-08683-y",
+   "title": "Prohormone cleavage prediction uncovers a non-incretin anti-obesity peptide",
+   "authors": [
+    "Laetitia Coassolo",
+    "Niels B. Danneskiold-Samsøe",
+    "Quennie Nguyen",
+    "Amanda Wiggenhorn",
+    "Meng Zhao",
+    "David Cheng-Hao Wang",
+    "David Toomer",
+    "Jameel Lone",
+    "Yichao Wei",
+    "Aayan Patel",
+    "Irene Liparulo",
+    "Deniz Kavi",
+    "Lianna W. Wat",
+    "Saranya Chidambaranathan Reghupaty",
+    "Julie Jae Kim",
+    "Tina Asemi",
+    "Ewa Bielczyk-Maczynska",
+    "Veronica L. Li",
+    "Maria Dolores Moya-Garzon",
+    "Nicole A. J. Krentz",
+    "Andreas Stahl",
+    "Danny Hung-Chieh Chou",
+    "Liqun Luo",
+    "Katrin J. Svensson"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-08683-y"
+  },
+  {
+   "task_id": "sgc-neuronal-nanotube-incidence",
+   "doi": "10.1038/s41586-025-09896-x",
+   "title": "Mitochondrial transfer from glia to neurons protects against peripheral neuropathy",
+   "authors": [
+    "Jing Xu",
+    "Yize Li",
+    "Charles Novak",
+    "Min Lee",
+    "Zihan Yan",
+    "Sangsu Bang",
+    "Aidan McGinnis",
+    "Sharat Chandra",
+    "Vivian Zhang",
+    "Wei He",
+    "Terry Lechler",
+    "Maria Pia Rodriguez Salazar",
+    "Cagla Eroglu",
+    "Matthew L. Becker",
+    "Dmitry Velmeshev",
+    "Richard E. Cheney",
+    "Ru-Rong Ji"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09896-x"
+  },
+  {
+   "task_id": "cell-state-ump-isotope-tracing-ranking",
+   "doi": "10.1038/s42255-026-01520-0",
+   "title": "Nitrogen metabolism profiling reveals cell state-specific pyrimidine synthesis pathway choice",
+   "authors": [
+    "Milan R. Savani",
+    "Bingbing Li",
+    "Bailey C. Smith",
+    "Wen Gu",
+    "Yi Xiao",
+    "Gerard Baquer",
+    "Tracey Shipman",
+    "Skyler S. Oken",
+    "Namya Manoj",
+    "Lauren G. Zacharias",
+    "Vinesh T. Puliyappadamba",
+    "Sylwia A. Stopka",
+    "Michael S. Regan",
+    "Michael M. Levitt",
+    "Charles K. Edgar",
+    "William H. Hicks",
+    "Soummitra Anand",
+    "Misty S. Martin-Sandoval",
+    "Rainah Winston",
+    "João S. Patrício",
+    "Xandria Johnson",
+    "Trevor S. Tippetts",
+    "Diana D. Shi",
+    "Andrew Lemoff",
+    "Timothy E. Richardson",
+    "Pascal O. Zinn",
+    "Ashley Solmonson",
+    "Thomas P. Mathews",
+    "Nathalie Y. R. Agar",
+    "Ralph J. DeBerardinis",
+    "Kalil G. Abdullah",
+    "Samuel K. McBrayer"
+   ],
+   "journal": "Nature Metabolism",
+   "url": "https://www.nature.com/articles/s42255-026-01520-0"
+  },
+  {
+   "task_id": "cancer-cell-ttfa-viability-ranking",
+   "doi": "10.1038/s42255-025-01410-x",
+   "title": "Pathway coessentiality mapping reveals complex II is required for de novo purine biosynthesis in acute myeloid leukaemia",
+   "authors": [
+    "Amy E. Stewart",
+    "Derek K. Zachman",
+    "Pol Castellano-Escuder",
+    "Lois M. Kelly",
+    "Ben Zolyomi",
+    "Michael D. I. Aiduk",
+    "Christopher D. Delaney",
+    "Ian C. Lock",
+    "Claudie Bosc",
+    "John Bradley",
+    "Shane T. Killarney",
+    "J. Darren Stuart",
+    "Paul A. Grimsrud",
+    "Olga R. Ilkayeva",
+    "Christopher B. Newgard",
+    "Navdeep S. Chandel",
+    "Alexandre Puissant",
+    "Kris C. Wood",
+    "Matthew D. Hirschey"
+   ],
+   "journal": "Nature Metabolism",
+   "url": "https://www.nature.com/articles/s42255-025-01410-x"
+  },
+  {
+   "task_id": "dna-polyploidization-route-fractions",
+   "doi": "10.1038/s41586-025-08986-0",
+   "title": "Multigenerational cell tracking of DNA replication and heritable DNA damage",
+   "authors": [
+    "Andreas Panagopoulos",
+    "Merula Stout",
+    "Sinan Kilic",
+    "Peter Leary",
+    "Julia Vornberger",
+    "Virginia Pasti",
+    "Antonio Galarreta",
+    "Aleksandra Lezaja",
+    "Kyra Kirschenbühler",
+    "Ralph Imhof",
+    "Hubert Rehrauer",
+    "Urs Ziegler",
+    "Matthias Altmeyer"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-08986-0"
+  },
+  {
+   "task_id": "ministeria-transcription-transcript-timecourse",
+   "doi": "10.1038/s41586-026-10748-5",
+   "title": "A unicellular relative links aggregative multicellularity to animal origins",
+   "authors": [
+    "Ruibao Li",
+    "Jennah E. Dharamshi",
+    "Kyle Kwok",
+    "Iñaki Ruiz-Trillo",
+    "Joseph P. Gerdt"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-026-10748-5"
+  },
+  {
+   "task_id": "ministeria-phagocytosis-transcript-timecourse",
+   "doi": "10.1038/s41586-026-10748-5",
+   "title": "A unicellular relative links aggregative multicellularity to animal origins",
+   "authors": [
+    "Ruibao Li",
+    "Jennah E. Dharamshi",
+    "Kyle Kwok",
+    "Iñaki Ruiz-Trillo",
+    "Joseph P. Gerdt"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-026-10748-5"
+  },
+  {
+   "task_id": "syoa-substrate-binding-ranking",
+   "doi": "10.1038/s41467-025-57129-6",
+   "title": "Structural insights into S-lignin O-demethylation via a rare class of heme peroxygenase enzymes",
+   "authors": [
+    "Alix C. Harlington",
+    "Tuhin Das",
+    "Keith E. Shearwin",
+    "Stephen G. Bell",
+    "Fiona Whelan"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-57129-6"
+  },
+  {
+   "task_id": "kemp-sequence-ranking",
+   "doi": "10.1038/s41586-025-09136-2",
+   "title": "Complete computational design of high-efficiency Kemp elimination enzymes",
+   "authors": [
+    "Dina Listov",
+    "Eva Vos",
+    "Gyula Hoffka",
+    "Shlomo Yakir Hoch",
+    "Andrej Berg",
+    "Shelly Hamer-Rogotner",
+    "Orly Dym",
+    "Shina Caroline Lynn Kamerlin",
+    "Sarel J. Fleishman"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09136-2"
+  },
+  {
+   "task_id": "kemp-sequence-thermal-stability",
+   "doi": "10.1038/s41586-025-09136-2",
+   "title": "Complete computational design of high-efficiency Kemp elimination enzymes",
+   "authors": [
+    "Dina Listov",
+    "Eva Vos",
+    "Gyula Hoffka",
+    "Shlomo Yakir Hoch",
+    "Andrej Berg",
+    "Shelly Hamer-Rogotner",
+    "Orly Dym",
+    "Shina Caroline Lynn Kamerlin",
+    "Sarel J. Fleishman"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09136-2"
+  },
+  {
+   "task_id": "p450bm3-quinoline-conversion-selectivity-ranking",
+   "doi": "10.1038/s41467-025-61773-3",
+   "title": "Regulating the N-oxidation selectivity of P450BM3 monooxygenases for N-heterocycles through computer-assisted structure-guided design",
+   "authors": [
+    "Liu Yang",
+    "Zhongji Pu",
+    "Jianping Wu",
+    "Xiaofeng Liu",
+    "Zhe Wang",
+    "Haoran Yu",
+    "Liuwei Wang",
+    "Yan Meng",
+    "Gang Xu",
+    "Lirong Yang",
+    "Wenlong Zheng"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-61773-3"
+  },
+  {
+   "task_id": "pes-h1-activity-thermal-ranking",
+   "doi": "10.1038/s41467-025-55944-5",
+   "title": "Tailoring industrial enzymes for thermostability and activity evolution by the machine learning-based iCASE strategy",
+   "authors": [
+    "Nan Zheng",
+    "Yongchao Cai",
+    "Zehua Zhang",
+    "Huimin Zhou",
+    "Yu Deng",
+    "Shuang Du",
+    "Mai Tu",
+    "Wei Fang",
+    "Xiaole Xia"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-55944-5"
+  },
+  {
+   "task_id": "mcba-cellfree-amidation-ranking",
+   "doi": "10.1038/s41467-024-55399-0",
+   "title": "Accelerated enzyme engineering by machine-learning guided cell-free expression",
+   "authors": [
+    "Grant M. Landwehr",
+    "Jonathan W. Bogart",
+    "Carol Magalhaes",
+    "Eric G. Hammarlund",
+    "Ashty S. Karim",
+    "Michael C. Jewett"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-024-55399-0"
+  },
+  {
+   "task_id": "diet-baseline-leukocyte-ranking",
+   "doi": "10.1038/s41586-026-10750-x",
+   "title": "Diet–microbiome synergy underlies obesity-associated immunotherapy efficacy",
+   "authors": [
+    "Lysanne Desharnais",
+    "Anikka Swaby",
+    "Meriem Messaoudene",
+    "Samuel Doré",
+    "Miranda W. Yu",
+    "Benoit Fiset",
+    "Valérie Breton",
+    "Mayra Ponce",
+    "Yongjia Hu",
+    "Liam Wilson",
+    "Mark Sorin",
+    "Ye Wang",
+    "Ken Dewar",
+    "Michael Pollak",
+    "Arielle Elkrief",
+    "Bertrand Routy",
+    "Logan A. Walsh",
+    "Daniela F. Quail"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-026-10750-x"
+  },
+  {
+   "task_id": "cortical-neuron-compound-dose-outgrowth",
+   "doi": "10.1038/s41586-025-09647-y",
+   "title": "Thiorphan reprograms neurons to promote functional recovery after spinal cord injury",
+   "authors": [
+    "E. A. van Niekerk",
+    "C. Marques de Freria",
+    "B. O. Mancarci",
+    "K. Groeniger",
+    "D. Kulinich",
+    "T. Riley",
+    "R. Kawaguchi",
+    "S. Okawa",
+    "T. Vokes",
+    "E. S. Rosenzweig",
+    "E. Sinopoulou",
+    "M. J. Castle",
+    "R. Huie",
+    "A. R. Ferguson",
+    "N. Kfoury-Beaumont",
+    "A. Khalessi",
+    "P. Pavlidis",
+    "M. H. Tuszynski"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09647-y"
+  },
+  {
+   "task_id": "erbb4-deletion-brain-histology",
+   "doi": "10.1038/s41586-026-10964-z",
+   "title": "Aberrant excitatory neuronal ERBB4 promotes Alzheimer’s disease pathology",
+   "authors": [
+    "Se Young Lee",
+    "Eunseok Park",
+    "Ha-Eun Lee",
+    "Seongbin Kim",
+    "Yeji Yeo",
+    "Juwon Park",
+    "Young-Jin Choi",
+    "Kiheon Lee",
+    "Ki-Jun Yoon",
+    "Sanghoon Park",
+    "Eunjoon Kim",
+    "Jae-Ick Kim",
+    "Won-Suk Chung"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-026-10964-z"
+  },
+  {
+   "task_id": "brp-sequence-acute-food-intake",
+   "doi": "10.1038/s41586-025-08683-y",
+   "title": "Prohormone cleavage prediction uncovers a non-incretin anti-obesity peptide",
+   "authors": [
+    "Laetitia Coassolo",
+    "Niels B. Danneskiold-Samsøe",
+    "Quennie Nguyen",
+    "Amanda Wiggenhorn",
+    "Meng Zhao",
+    "David Cheng-Hao Wang",
+    "David Toomer",
+    "Jameel Lone",
+    "Yichao Wei",
+    "Aayan Patel",
+    "Irene Liparulo",
+    "Deniz Kavi",
+    "Lianna W. Wat",
+    "Saranya Chidambaranathan Reghupaty",
+    "Julie Jae Kim",
+    "Tina Asemi",
+    "Ewa Bielczyk-Maczynska",
+    "Veronica L. Li",
+    "Maria Dolores Moya-Garzon",
+    "Nicole A. J. Krentz",
+    "Andreas Stahl",
+    "Danny Hung-Chieh Chou",
+    "Liqun Luo",
+    "Katrin J. Svensson"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-08683-y"
+  },
+  {
+   "task_id": "picalm-lipid-class-abundance",
+   "doi": "10.1038/s41586-025-09486-x",
+   "title": "PICALM Alzheimer’s risk allele causes aberrant lipid droplets in microglia",
+   "authors": [
+    "Alena Kozlova",
+    "Siwei Zhang",
+    "Ari Sudwarts",
+    "Hanwen Zhang",
+    "Stanislau Smirnou",
+    "Seul Kee Byeon",
+    "Christina Thapa",
+    "Xiaotong Sun",
+    "Kimberley Stephenson",
+    "Xiaojie Zhao",
+    "Brendan Jamison",
+    "Moorthi Ponnusamy",
+    "Xin He",
+    "Julie A. Schneider",
+    "Akhilesh Pandey",
+    "David A. Bennett",
+    "Zhiping P. Pang",
+    "Alan R. Sanders",
+    "Hugo J. Bellen",
+    "Gopal Thinakaran",
+    "Jubao Duan"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09486-x"
+  },
+  {
+   "task_id": "pancreatic-kras-lipogenic-transcripts",
+   "doi": "10.1038/s41586-025-08917-z",
+   "title": "Targeting PIKfyve-driven lipid metabolism in pancreatic cancer",
+   "authors": [
+    "Caleb Cheng",
+    "Jing Hu",
+    "Rahul Mannan",
+    "Tongchen He",
+    "Rupam Bhattacharyya",
+    "Brian Magnuson",
+    "Jasmine P. Wisniewski",
+    "Sydney Peters",
+    "Saadia A. Karim",
+    "David J. MacLean",
+    "Hüseyin Karabürk",
+    "Li Zhang",
+    "Nicholas J. Rossiter",
+    "Yang Zheng",
+    "Lanbo Xiao",
+    "Chungen Li",
+    "Dominik Awad",
+    "Somnath Mahapatra",
+    "Yi Bao",
+    "Yuping Zhang",
+    "Xuhong Cao",
+    "Zhen Wang",
+    "Rohit Mehra",
+    "Pietro Morlacchi",
+    "Vaibhav Sahai",
+    "Marina Pasca di Magliano",
+    "Yatrik M. Shah",
+    "Lois S. Weisman",
+    "Jennifer P. Morton",
+    "Ke Ding",
+    "Yuanyuan Qiao",
+    "Costas A. Lyssiotis",
+    "Arul M. Chinnaiyan"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-08917-z"
+  },
+  {
+   "task_id": "human-tumour-oxidized-lipid-response",
+   "doi": "10.1038/s41586-025-08974-4",
+   "title": "Activation of lysosomal iron triggers ferroptosis in cancer",
+   "authors": [
+    "Tatiana Cañeque",
+    "Leeroy Baron",
+    "Sebastian Müller",
+    "Alanis Carmona",
+    "Ludovic Colombeau",
+    "Antoine Versini",
+    "Stéphanie Solier",
+    "Christine Gaillet",
+    "Fabien Sindikubwabo",
+    "Julio L. Sampaio",
+    "Marie Sabatier",
+    "Eikan Mishima",
+    "Armel Picard-Bernes",
+    "Laurène Syx",
+    "Nicolas Servant",
+    "Bérangère Lombard",
+    "Damarys Loew",
+    "Jiashuo Zheng",
+    "Bettina Proneth",
+    "Leishemba K. Thoidingjam",
+    "Laurence Grimaud",
+    "Cameron S. Fraser",
+    "Krystina J. Szylo",
+    "Emma Der Kazarian",
+    "Caroline Bonnet",
+    "Emmanuelle Charafe-Jauffret",
+    "Christophe Ginestier",
+    "Patricia Santofimia-Castaño",
+    "Matias Estaras",
+    "Nelson Dusetti",
+    "Juan Lucio Iovanna",
+    "Antonio Sa Cunha",
+    "Gabriella Pittau",
+    "Pascal Hammel",
+    "Dimitri Tzanis",
+    "Sylvie Bonvalot",
+    "Sarah Watson",
+    "Vincent Gandon",
+    "Aditya Upadhyay",
+    "Derek A. Pratt",
+    "Florêncio Porto Freitas",
+    "José Pedro Friedmann Angeli",
+    "Brent R. Stockwell",
+    "Marcus Conrad",
+    "Jessalyn M. Ubellacker",
+    "Raphaël Rodriguez"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-08974-4"
+  },
+  {
+   "task_id": "cancer-cell-atra-ic50-ranking",
+   "doi": "10.1038/s41467-025-57185-y",
+   "title": "Bone morphogenetic protein (BMP) signaling determines neuroblastoma cell fate and sensitivity to retinoic acid",
+   "authors": [
+    "Min Pan",
+    "Yinwen Zhang",
+    "William C. Wright",
+    "Xueying Liu",
+    "Barbara Passaia",
+    "Duane Currier",
+    "Jonathan Low",
+    "Richard H. Chapple",
+    "Jacob A. Steele",
+    "Jon P. Connelly",
+    "Bensheng Ju",
+    "Emily Plyler",
+    "Meifen Lu",
+    "Allister J. Loughran",
+    "Lei Yang",
+    "Brian J. Abraham",
+    "Shondra M. Pruett-Miller",
+    "Burgess Freeman",
+    "George E. Campbell",
+    "Michael A. Dyer",
+    "Taosheng Chen",
+    "Elizabeth Stewart",
+    "Selene Koo",
+    "Heather Sheppard",
+    "John Easton",
+    "Paul Geeleher"
+   ],
+   "journal": "Nature Communications",
+   "url": "https://www.nature.com/articles/s41467-025-57185-y"
+  },
+  {
+   "task_id": "pooled-cell-drug-representation-ranking",
+   "doi": "10.1038/s43018-026-01130-5",
+   "title": "The GENEVA platform models tumor mosaicism to reveal variations of responses to KRAS inhibitors and identify improved drug combinations",
+   "authors": [
+    "Johnny X. Yu",
+    "Jung Min Suh",
+    "Katerina D. Popova",
+    "Kristle Garcia",
+    "Tanvi Joshi",
+    "Bruce Culbertson",
+    "Jessica B. Spinelli",
+    "Vishvak Subramanyam",
+    "Kevin Lou",
+    "Trey Charbonneau",
+    "Kevan M. Shokat",
+    "Jonathan Weissman",
+    "Hani Goodarzi"
+   ],
+   "journal": "Nature Cancer",
+   "url": "https://www.nature.com/articles/s43018-026-01130-5"
+  },
+  {
+   "task_id": "grpr-yap1-nuclear-localization",
+   "doi": "10.1038/s41586-025-09111-x",
+   "title": "Targeting GRPR for sex hormone-dependent cancer after loss of E-cadherin",
+   "authors": [
+    "Jérémy H. Raymond",
+    "Zackie Aktary",
+    "Marie Pouteaux",
+    "Valérie Petit",
+    "Flavie Luciani",
+    "Maria Wehbe",
+    "Patrick Gizzi",
+    "Claire Bourban",
+    "Didier Decaudin",
+    "Fariba Nemati",
+    "Igor Martianov",
+    "Irwin Davidson",
+    "Catherine-Laure Tomasetto",
+    "Richard M. White",
+    "Florence Mahuteau-Betzer",
+    "Béatrice Vergier",
+    "Lionel Larue",
+    "Véronique Delmas"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09111-x"
+  },
+  {
+   "task_id": "grpr-melanoma-matrigel-invasion",
+   "doi": "10.1038/s41586-025-09111-x",
+   "title": "Targeting GRPR for sex hormone-dependent cancer after loss of E-cadherin",
+   "authors": [
+    "Jérémy H. Raymond",
+    "Zackie Aktary",
+    "Marie Pouteaux",
+    "Valérie Petit",
+    "Flavie Luciani",
+    "Maria Wehbe",
+    "Patrick Gizzi",
+    "Claire Bourban",
+    "Didier Decaudin",
+    "Fariba Nemati",
+    "Igor Martianov",
+    "Irwin Davidson",
+    "Catherine-Laure Tomasetto",
+    "Richard M. White",
+    "Florence Mahuteau-Betzer",
+    "Béatrice Vergier",
+    "Lionel Larue",
+    "Véronique Delmas"
+   ],
+   "journal": "Nature",
+   "url": "https://www.nature.com/articles/s41586-025-09111-x"
+  },
+  {
+   "task_id": "ly2090314-cell-growth-inhibition-ranking",
+   "doi": "10.1038/s43018-025-01093-z",
+   "title": "Targeting β-catenin degradation with GSK3β inhibitors induces cell death in acute lymphoblastic leukemia",
+   "authors": [
+    "Kadriye Nehir Cosgun",
+    "Huda Jumaa",
+    "Mark E. Robinson",
+    "Zhangliang Cheng",
+    "Salim Oulghazi",
+    "Kohei Kume",
+    "David Fonseca Arce",
+    "Nikol Agadzhanian",
+    "Klaus M. Kistner",
+    "Etienne Leveille",
+    "Elsa Drivet",
+    "Fang Yu",
+    "Zhijian Qian",
+    "Joo Y. Song",
+    "Wing-Chung Chan",
+    "Liang Xu",
+    "Gang Xiao",
+    "M. Mark Taketo",
+    "Shalin Kothari",
+    "Matthew S. Davids",
+    "Hilde Schjerven",
+    "Julia Jellusova",
+    "Markus Müschen"
+   ],
+   "journal": "Nature Cancer",
+   "url": "https://www.nature.com/articles/s43018-025-01093-z"
+  }
+ ]
+}

eval.yaml

@@ -0,0 +1,12 @@
+name: Science Prediction Universe
+description: >
+  Science Prediction Universe (SPU) tests whether AI agents can predict the outcomes of
+  real wet-lab experiments. Each of the 100 public tasks asks the agent to rank
+  experimental candidates by a measured outcome, and rankings are scored by
+  Spearman correlation. Report the number of scored tasks, model version,
+  reasoning effort and harness with every result.
+evaluation_framework: harbor
+tasks:
+  - id: public100-scored-mean-spearman-v1
+    config: default
+    split: test

paper/Science_Prediction_Universe.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7f70c813a686f31e1799079931cf531a010e536d2d5fab02521bf36ec30a0dd
+size 3971574

task_cards.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab80ef74a4226fedab9512372dcbee8f5b905e81350ff88b87fcd3b993f9e71b
+size 30582

tasks/agcute-bending-endpoints-ranking/environment/Dockerfile

@@ -0,0 +1,7 @@
+FROM python@sha256:78387bc3881b8273120a12ebe6c1ab22b018ccc2c9adf565ae1ac9b536e184ea
+ENV PYTHONDONTWRITEBYTECODE=1 PYTHONUNBUFFERED=1
+RUN useradd -m -u 1000 -s /bin/bash solver && mkdir -p /logs/artifacts && chmod 777 /logs/artifacts
+WORKDIR /workspace
+COPY public/ /workspace/public/
+ENV OPENBLAS_NUM_THREADS=8 OMP_NUM_THREADS=8 MKL_NUM_THREADS=8
+USER solver

tasks/agcute-bending-endpoints-ranking/environment/public/contract.json

@@ -0,0 +1,98 @@
+{
+  "task_id": "agcute-bending-endpoints-ranking",
+  "task_version": "1.0.0",
+  "groups": [
+    {
+      "group_id": "max_stress",
+      "label": "Alloy composition series: bending stress",
+      "item_ids": [
+        "exp_65960918c5e1",
+        "exp_0b361eedc092",
+        "exp_0a801637748b",
+        "exp_39fd1eed6b1c",
+        "exp_365507fb236c",
+        "exp_3c3c8522da9d",
+        "exp_fc4984656462",
+        "exp_3fed9f8f07c9",
+        "exp_aaa9e2c6f5b9"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.5,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "Maximum bending stress",
+        "unit": "MPa",
+        "definition_en": "Largest recorded engineering bending stress in each curve."
+      },
+      "uncertainty_definition": "Reliable independent uncertainty is not available for this panel. Unequal reported values are compared directly; uncertainty is not assumed to be zero."
+    },
+    {
+      "group_id": "strain_at_max",
+      "label": "Alloy composition series: peak-stress strain",
+      "item_ids": [
+        "exp_54b8fc8b3571",
+        "exp_3aa8fedf978b",
+        "exp_2d9d2ad06f6e",
+        "exp_8cfb0d14a259",
+        "exp_d53afffea8ad",
+        "exp_69a2a5342ec6",
+        "exp_eb05cc2a9dc8",
+        "exp_7b741589c10a",
+        "exp_4369fc61d66d"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.5,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "Engineering strain at maximum bending stress",
+        "unit": "%",
+        "definition_en": "Engineering strain at the same sampled point as the largest bending stress, using the earliest occurrence if maxima repeat."
+      },
+      "uncertainty_definition": "Reliable independent uncertainty is not available for this panel. Unequal reported values are compared directly; uncertainty is not assumed to be zero."
+    }
+  ],
+  "scoring_method": "spearman",
+  "primary_score": "task_spearman",
+  "spearman_range": [
+    -1,
+    1
+  ],
+  "spearman_tie_method": "average",
+  "spearman_undefined_policy": "task_error",
+  "spearman_aggregation": "Pearson correlation of true average ranks and submitted positions, separately within each group, then fixed group-weighted sum; do not pool groups or clip negative groups.",
+  "secondary_scoring_method": "equal_pair_inversion_random95",
+  "ranking_reward_range": [
+    0,
+    1
+  ],
+  "random_quantile": 0.95,
+  "unknown_uncertainty_policy": "use_reported_value",
+  "pair_rule": "For each unordered pair, count it iff abs(mu_i-mu_j)>k*s_ij+epsilon for known uncertainty, or >epsilon for unknown uncertainty. Each counted pair has equal weight within its group. Unknown uncertainty is not zero uncertainty.",
+  "ranking_reward_formula": "Validate finite 0.5<=q<1 and 0<=A<=1; if A==1 return 1; else if A<=q+1e-12 return 0; otherwise clamp((A-q)/(1-q),0,1). A is the fixed weighted sum of group pair accuracies; q is the whole-task random 95th percentile.",
+  "format_errors": {
+    "whole_task": "Core=-1; auxiliary=0 for malformed/oversized/non-UTF8 JSON, duplicate object keys, wrong/missing identity, invalid top-level/rankings, unknown or unassignable group.",
+    "known_group": "Missing/duplicate known group or invalid item permutation: raw group Spearman=null, aggregation value=-1, group pair accuracy=0; preserve group weight and apply auxiliary threshold after aggregation."
+  },
+  "max_submission_bytes": 1048576,
+  "minimum_conditions_per_group": 2,
+  "time_limit_sec": 28800,
+  "solver_resources": {
+    "cpus": 8,
+    "memory_mb": 16384,
+    "gpus": 1,
+    "gpu_types": [
+      "H20"
+    ]
+  },
+  "submission_path": "/logs/artifacts/rankings.json",
+  "report_path": "/logs/artifacts/report.md",
+  "sources_path": "/logs/artifacts/sources.json",
+  "analysis_path": "/logs/artifacts/analysis/",
+  "forbidden_result_scope": "Experimental bending stress-strain, flexural failure, toughness or equivalent mechanical rankings for Ag-Cu-Te-Se-S chalcogenide alloy composition and vacancy series are same-type results and prohibited, including related composition or test-protocol variants. General solid mechanics and independent chemical or crystallographic basic facts outside these result panels are allowed."
+}

tasks/agcute-bending-endpoints-ranking/environment/public/experiment.md

@@ -0,0 +1,30 @@
+# Experimental definition
+
+Compare nine nominal alloy compositions produced by melting, annealing, quenching and spark plasma sintering. Rank the maximum recorded bending stress and the engineering strain at that maximum in separate groups.
+
+## Common conditions
+
+| Field | Setting |
+| --- | --- |
+| raw_materials | `Ag (99.99%), Cu (99.99%), Te (99.999%), Se (99.999%) and S (99.999%) in the listed nominal stoichiometry; omit Se and S for AgCuTe` |
+| melting_protocol | `Vacuum-sealed pyrolytic-carbon-lined quartz tube; heat to 773 K over 12 h, then to 1273 K over 5 h; hold 1273 K for 48 h` |
+| anneal_quench | `Cool to 873 K over 4 h, hold 3 days, quench in ice water; cool in liquid nitrogen and grind in an agate mortar` |
+| sintering | `Spark plasma sintering in vacuum in a 12.7 mm graphite die at 773 K for 5 min under approximately 4.0 kN applied load` |
+| bending_geometry_mm | `1 x 0.5 x 10 mm; support span not specified` |
+| bending_measurement | `Three-point engineering bending stress-strain; DMA 850 (TA Instruments), room temperature, loading rate 0.1 mm/min` |
+
+## Procedure and measurement
+
+Follow the stated synthesis and three-point bending test. For each original engineering stress-strain trace, use its largest reported bending-stress sample and the strain at that same sample; if the maximum is repeated, use the earliest recorded occurrence. Negative pre-load samples are retained in the trace. Trace endpoints are not treated as fracture measurements.
+
+- `max_stress` — Maximum bending stress (MPa): Largest recorded engineering bending stress in each curve.
+- `strain_at_max` — Engineering strain at maximum bending stress (%): Engineering strain at the same sampled point as the largest bending stress, using the earliest occurrence if maxima repeat.
+
+## Item settings
+
+Expand each item in targets.csv or protocol.json together with the common conditions. Every selected item is listed using the same fields.
+
+## Source limitations relevant to the definition
+
+- The bending support span, independent specimen count and specimen-to-specimen uncertainty are not reported for this curve panel.
+- These are curve maxima and their associated strains, not an inferred fracture strain or an independently measured toughness.

tasks/agcute-bending-endpoints-ranking/environment/public/information_policy.md

@@ -0,0 +1,11 @@
+# Information policy
+
+The original study, all its supplementary materials (including methods), source data, related reporting, summaries, mirrors and caches are prohibited. The same restriction covers external values, rankings, qualitative comparisons, equivalent endpoints and predictors fitted to target or same-type experimental results. Experimental bending stress-strain, flexural failure, toughness or equivalent mechanical rankings for Ag-Cu-Te-Se-S chalcogenide alloy composition and vacancy series are same-type results and prohibited, including related composition or test-protocol variants. General solid mechanics and independent chemical or crystallographic basic facts outside these result panels are allowed.
+
+Allowed examples:
+- Independent general theory of the stated measurement and material chemistry; calculations based on the supplied experimental settings.
+
+Forbidden examples:
+- The source study or its supplementary material; equivalent experimental screening values or rankings; predictors trained or calibrated on those target or same-type results.
+
+Use audited tools for all searches, downloads and code-initiated network requests. Check both requests and returned content before use. Do not circumvent access controls or ask another model/service for forbidden results. If a permitted query unexpectedly exposes prohibited external results, immediately stop prediction and further search; record only the triggering step, source and stopping reason in report.md without restating results. Do not retrieve or reconstruct specific experimental results from your own model memory; briefly disclose a noticed recollection in your report and continue using permitted inputs and reasoning.

tasks/agcute-bending-endpoints-ranking/environment/public/protocol.json

@@ -0,0 +1,193 @@
+{
+  "common_conditions": {
+    "raw_materials": "Ag (99.99%), Cu (99.99%), Te (99.999%), Se (99.999%) and S (99.999%) in the listed nominal stoichiometry; omit Se and S for AgCuTe",
+    "melting_protocol": "Vacuum-sealed pyrolytic-carbon-lined quartz tube; heat to 773 K over 12 h, then to 1273 K over 5 h; hold 1273 K for 48 h",
+    "anneal_quench": "Cool to 873 K over 4 h, hold 3 days, quench in ice water; cool in liquid nitrogen and grind in an agate mortar",
+    "sintering": "Spark plasma sintering in vacuum in a 12.7 mm graphite die at 773 K for 5 min under approximately 4.0 kN applied load",
+    "bending_geometry_mm": "1 x 0.5 x 10 mm; support span not specified",
+    "bending_measurement": "Three-point engineering bending stress-strain; DMA 850 (TA Instruments), room temperature, loading rate 0.1 mm/min"
+  },
+  "measurement": {
+    "name": "Group-specific measured endpoint",
+    "unit": "group-specific",
+    "definition_en": "Use the endpoint and unit specified for each comparison group."
+  },
+  "measurements_by_group": {
+    "max_stress": {
+      "name": "Maximum bending stress",
+      "unit": "MPa",
+      "definition_en": "Largest recorded engineering bending stress in each curve."
+    },
+    "strain_at_max": {
+      "name": "Engineering strain at maximum bending stress",
+      "unit": "%",
+      "definition_en": "Engineering strain at the same sampled point as the largest bending stress, using the earliest occurrence if maxima repeat."
+    }
+  },
+  "measurement_rule": "Use measurements_by_group for each group. The measurement field is a default summary and does not override a group-specific endpoint.",
+  "inheritance_rule": "common_conditions apply unless the same field is explicitly overridden by the item; item settings are literal experimental facts, not predicted outcomes.",
+  "items": {
+    "exp_65960918c5e1": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.85,
+        0.1,
+        0.05
+      ]
+    },
+    "exp_0b361eedc092": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.8,
+        0.1,
+        0.1
+      ]
+    },
+    "exp_0a801637748b": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.78,
+        0.1,
+        0.12
+      ]
+    },
+    "exp_39fd1eed6b1c": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.7,
+        0.1,
+        0.2
+      ]
+    },
+    "exp_365507fb236c": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        0.998,
+        0.998,
+        0.8,
+        0.1,
+        0.1
+      ]
+    },
+    "exp_3c3c8522da9d": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.82,
+        0.1,
+        0.08
+      ]
+    },
+    "exp_fc4984656462": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.75,
+        0.1,
+        0.15
+      ]
+    },
+    "exp_3fed9f8f07c9": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        1,
+        0,
+        0
+      ]
+    },
+    "exp_aaa9e2c6f5b9": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.9,
+        0.1,
+        0
+      ]
+    },
+    "exp_54b8fc8b3571": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.85,
+        0.1,
+        0.05
+      ]
+    },
+    "exp_3aa8fedf978b": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        0.998,
+        0.998,
+        0.8,
+        0.1,
+        0.1
+      ]
+    },
+    "exp_2d9d2ad06f6e": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.8,
+        0.1,
+        0.1
+      ]
+    },
+    "exp_8cfb0d14a259": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.78,
+        0.1,
+        0.12
+      ]
+    },
+    "exp_d53afffea8ad": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.9,
+        0.1,
+        0
+      ]
+    },
+    "exp_69a2a5342ec6": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        1,
+        0,
+        0
+      ]
+    },
+    "exp_eb05cc2a9dc8": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.7,
+        0.1,
+        0.2
+      ]
+    },
+    "exp_7b741589c10a": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.82,
+        0.1,
+        0.08
+      ]
+    },
+    "exp_4369fc61d66d": {
+      "nominal_atomic_ratio_Ag_Cu_Te_Se_S": [
+        1,
+        1,
+        0.75,
+        0.1,
+        0.15
+      ]
+    }
+  }
+}

tasks/agcute-bending-endpoints-ranking/environment/public/resources.md

@@ -0,0 +1,3 @@
+# Resources
+
+You have 8 CPU cores, 16,384 MiB system RAM and one NVIDIA H20 GPU, shared by all local processes. Remote model APIs, databases and services do not consume the local compute quota; their calls and waits consume the same 28,800-second wall-clock budget. Record service identity, requests and usage. All compute and network activity must follow the information policy. The 8-hour clock starts at first delivery of the task inputs; retries do not reset it.

tasks/agcute-bending-endpoints-ranking/environment/public/targets.csv

@@ -0,0 +1,19 @@
+group_id,item_id,settings_ref,nominal_atomic_ratio_Ag_Cu_Te_Se_S
+max_stress,exp_65960918c5e1,protocol.json#exp_65960918c5e1,"[1, 1, 0.85, 0.1, 0.05]"
+max_stress,exp_0b361eedc092,protocol.json#exp_0b361eedc092,"[1, 1, 0.8, 0.1, 0.1]"
+max_stress,exp_0a801637748b,protocol.json#exp_0a801637748b,"[1, 1, 0.78, 0.1, 0.12]"
+max_stress,exp_39fd1eed6b1c,protocol.json#exp_39fd1eed6b1c,"[1, 1, 0.7, 0.1, 0.2]"
+max_stress,exp_365507fb236c,protocol.json#exp_365507fb236c,"[0.998, 0.998, 0.8, 0.1, 0.1]"
+max_stress,exp_3c3c8522da9d,protocol.json#exp_3c3c8522da9d,"[1, 1, 0.82, 0.1, 0.08]"
+max_stress,exp_fc4984656462,protocol.json#exp_fc4984656462,"[1, 1, 0.75, 0.1, 0.15]"
+max_stress,exp_3fed9f8f07c9,protocol.json#exp_3fed9f8f07c9,"[1, 1, 1, 0, 0]"
+max_stress,exp_aaa9e2c6f5b9,protocol.json#exp_aaa9e2c6f5b9,"[1, 1, 0.9, 0.1, 0]"
+strain_at_max,exp_54b8fc8b3571,protocol.json#exp_54b8fc8b3571,"[1, 1, 0.85, 0.1, 0.05]"
+strain_at_max,exp_3aa8fedf978b,protocol.json#exp_3aa8fedf978b,"[0.998, 0.998, 0.8, 0.1, 0.1]"
+strain_at_max,exp_2d9d2ad06f6e,protocol.json#exp_2d9d2ad06f6e,"[1, 1, 0.8, 0.1, 0.1]"
+strain_at_max,exp_8cfb0d14a259,protocol.json#exp_8cfb0d14a259,"[1, 1, 0.78, 0.1, 0.12]"
+strain_at_max,exp_d53afffea8ad,protocol.json#exp_d53afffea8ad,"[1, 1, 0.9, 0.1, 0]"
+strain_at_max,exp_69a2a5342ec6,protocol.json#exp_69a2a5342ec6,"[1, 1, 1, 0, 0]"
+strain_at_max,exp_eb05cc2a9dc8,protocol.json#exp_eb05cc2a9dc8,"[1, 1, 0.7, 0.1, 0.2]"
+strain_at_max,exp_7b741589c10a,protocol.json#exp_7b741589c10a,"[1, 1, 0.82, 0.1, 0.08]"
+strain_at_max,exp_4369fc61d66d,protocol.json#exp_4369fc61d66d,"[1, 1, 0.75, 0.1, 0.15]"

tasks/agcute-bending-endpoints-ranking/instruction.md

@@ -0,0 +1,203 @@
+You are a general-purpose AI scientist tasked with predicting experimental outcomes across scientific disciplines. Your goal is to produce the most accurate, scientifically grounded predictions possible from the information provided and any permitted information you can obtain. For this task, predict only the ordering of experimental outcomes within each specified comparison group. Absolute outcome values are not required.
+
+You may use all computational tools and search resources available in your environment, including web search, scientific literature, databases, code execution, statistical modeling, machine learning, and numerical simulation. Select and combine methods appropriate to the task, and actively perform calculations when they can improve your predictions. Proactively check the available CPUs, GPUs, and software, and identify key prediction problems they could help address; whether to use a GPU should depend on the suitability of the method. For scientifically grounded methods that may be useful but whose benefits are uncertain, prioritize small-scale trial calculations before deciding whether to expand the computation. You may download any software, code, models, and data you consider necessary from the internet, subject to the following restrictions: do not search for, access, or use the original study, its supplementary materials, or data that can directly reveal the experimental results, and do not use the results of the target experiments or experiments of the same type.
+
+Ground your analysis in the experimental setup, measurement definitions, controls, and relevant scientific mechanisms. Gather supporting evidence from permitted sources, assess its applicability to the specific experiments, and evaluate alternative explanations. Check assumptions, units, and consistency, and refine your predictions as new permitted evidence becomes available. Use general scientific knowledge to make new predictions. Do not retrieve, reconstruct, or use specific experimental results or rankings from your model memory or training data. After forming an initial ranking, continue to look for and actually try methods that could improve key judgments within the time budget, prioritizing resources for the most uncertain comparisons.
+
+Clearly distinguish experimental observations, computational results, and scientific inference. Assess uncertainty and methodological limitations. Do not fabricate sources, data, tool outputs, or calculations, and do not present computational proxies as experimentally measured values.
+
+Follow the task-specific instructions and output format exactly. When explanations are permitted, provide a concise summary of your methods, key computational results, and supporting rationale. Keep supporting analysis separate from any final submission that requires a restricted format.
+
+You have a strict time limit of 8 hours (28,800 seconds). Submit the required files before the deadline.
+
+# Bending stress and peak-stress strain of Ag-Cu-Te-Se-S alloys
+
+## Prediction question
+
+Compare nine nominal alloy compositions produced by melting, annealing, quenching and spark plasma sintering. Rank the maximum recorded bending stress and the engineering strain at that maximum in separate groups.
+
+Rank each comparison group from lowest to highest using that group's specified endpoint and unit. Submit rankings only; absolute values are not required.
+
+## Experimental definition
+
+Read `/workspace/public/experiment.md`, `protocol.json` and `targets.csv`. The common conditions in protocol.json apply to every item unless that item's setting explicitly overrides the same field. Expand both parts together. Public sequences, structures and other supplied assets are part of the definition. Do not obtain missing details from the prohibited original study.
+
+Follow the stated synthesis and three-point bending test. For each original engineering stress-strain trace, use its largest reported bending-stress sample and the strain at that same sample; if the maximum is repeated, use the earliest recorded occurrence. Negative pre-load samples are retained in the trace. Trace endpoints are not treated as fracture measurements.
+
+Outcome definitions by group:
+
+- `max_stress` — Maximum bending stress (MPa): Largest recorded engineering bending stress in each curve.
+- `strain_at_max` — Engineering strain at maximum bending stress (%): Engineering strain at the same sampled point as the largest bending stress, using the earliest occurrence if maxima repeat.
+
+## Comparison groups
+
+- `max_stress`: Alloy composition series: bending stress; 9 items; weight 0.5; ascending Maximum bending stress (MPa).
+- `strain_at_max`: Alloy composition series: peak-stress strain; 9 items; weight 0.5; ascending Engineering strain at maximum bending stress (%).
+
+## Allowed and prohibited information
+
+The original study, all its supplementary materials (including methods), source data, related reporting, summaries, mirrors and caches are prohibited. The same restriction covers external values, rankings, qualitative comparisons, equivalent endpoints and predictors fitted to target or same-type experimental results. Experimental bending stress-strain, flexural failure, toughness or equivalent mechanical rankings for Ag-Cu-Te-Se-S chalcogenide alloy composition and vacancy series are same-type results and prohibited, including related composition or test-protocol variants. General solid mechanics and independent chemical or crystallographic basic facts outside these result panels are allowed.
+
+Allowed examples:
+- Independent general theory of the stated measurement and material chemistry; calculations based on the supplied experimental settings.
+
+Forbidden examples:
+- The source study or its supplementary material; equivalent experimental screening values or rankings; predictors trained or calibrated on those target or same-type results.
+
+Use audited tools for all searches, downloads and code-initiated network requests. Check both requests and returned content before use. Do not circumvent access controls or ask another model/service for forbidden results. If a permitted query unexpectedly exposes prohibited external results, immediately stop prediction and further search; record only the triggering step, source and stopping reason in report.md without restating results. Do not retrieve or reconstruct specific experimental results from your own model memory; briefly disclose a noticed recollection in your report and continue using permitted inputs and reasoning.
+
+## Tools, resources and deadline
+
+You have 8 CPU cores, 16,384 MiB system RAM and one NVIDIA H20 GPU, shared by all local processes. Remote model APIs, databases and services do not consume the local compute quota; their calls and waits consume the same 28,800-second wall-clock budget. Record service identity, requests and usage. All compute and network activity must follow the information policy. The 8-hour clock starts at first delivery of the task inputs; retries do not reset it.
+
+## Submission
+
+Write UTF-8 JSON to `/logs/artifacts/rankings.json` (at most 1 MiB, unique object keys). Include the exact task identity and exactly one entry per group; each order must contain every assigned item once, from lowest to highest. The following is a format example in public display order, not a predicted answer:
+
+```json
+{
+  "task_id": "agcute-bending-endpoints-ranking",
+  "task_version": "1.0.0",
+  "rankings": [
+    {
+      "group_id": "max_stress",
+      "order": [
+        "exp_65960918c5e1",
+        "exp_0b361eedc092",
+        "exp_0a801637748b",
+        "exp_39fd1eed6b1c",
+        "exp_365507fb236c",
+        "exp_3c3c8522da9d",
+        "exp_fc4984656462",
+        "exp_3fed9f8f07c9",
+        "exp_aaa9e2c6f5b9"
+      ]
+    },
+    {
+      "group_id": "strain_at_max",
+      "order": [
+        "exp_54b8fc8b3571",
+        "exp_3aa8fedf978b",
+        "exp_2d9d2ad06f6e",
+        "exp_8cfb0d14a259",
+        "exp_d53afffea8ad",
+        "exp_69a2a5342ec6",
+        "exp_eb05cc2a9dc8",
+        "exp_7b741589c10a",
+        "exp_4369fc61d66d"
+      ]
+    }
+  ]
+}
+```
+
+Write an English account of scientific reasoning, calculations, assumptions and uncertainty to `/logs/artifacts/report.md`; source records to `/logs/artifacts/sources.json`; supporting scripts and calculations to `/logs/artifacts/analysis/`. The evaluator uses the last submission received before the deadline. Empty or missing submissions receive core -1 and auxiliary 0.
+
+## Ranking scores
+
+The primary score is standard Spearman rank correlation, from -1 to 1, computed separately in each group and combined with fixed group weights. Equal measured values receive average true ranks; predictions remain strict permutations. Correct/reverse rankings score 1/-1 when true values have no ties; with true ties the best strict ranking can score less than 1. Random complete rankings have expected correlation 0.
+
+The auxiliary score counts each decidable pair equally within its group, combines group accuracies using the same weights, then applies one whole-task random 95th-percentile threshold. Known uncertainty uses the declared difference-error rule; unknown uncertainty uses reported values with epsilon and does not exclude unequal results. The threshold is hidden. Core and auxiliary scores are recorded separately; neither meaning of zero proves the absence of predictive ability. Detailed algorithms and all format penalties are in the following authoritative public contract, from which the companion files are generated.
+
+<!-- public-contract -->
+```json
+{
+  "task_id": "agcute-bending-endpoints-ranking",
+  "task_version": "1.0.0",
+  "groups": [
+    {
+      "group_id": "max_stress",
+      "label": "Alloy composition series: bending stress",
+      "item_ids": [
+        "exp_65960918c5e1",
+        "exp_0b361eedc092",
+        "exp_0a801637748b",
+        "exp_39fd1eed6b1c",
+        "exp_365507fb236c",
+        "exp_3c3c8522da9d",
+        "exp_fc4984656462",
+        "exp_3fed9f8f07c9",
+        "exp_aaa9e2c6f5b9"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.5,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "Maximum bending stress",
+        "unit": "MPa",
+        "definition_en": "Largest recorded engineering bending stress in each curve."
+      },
+      "uncertainty_definition": "Reliable independent uncertainty is not available for this panel. Unequal reported values are compared directly; uncertainty is not assumed to be zero."
+    },
+    {
+      "group_id": "strain_at_max",
+      "label": "Alloy composition series: peak-stress strain",
+      "item_ids": [
+        "exp_54b8fc8b3571",
+        "exp_3aa8fedf978b",
+        "exp_2d9d2ad06f6e",
+        "exp_8cfb0d14a259",
+        "exp_d53afffea8ad",
+        "exp_69a2a5342ec6",
+        "exp_eb05cc2a9dc8",
+        "exp_7b741589c10a",
+        "exp_4369fc61d66d"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.5,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "Engineering strain at maximum bending stress",
+        "unit": "%",
+        "definition_en": "Engineering strain at the same sampled point as the largest bending stress, using the earliest occurrence if maxima repeat."
+      },
+      "uncertainty_definition": "Reliable independent uncertainty is not available for this panel. Unequal reported values are compared directly; uncertainty is not assumed to be zero."
+    }
+  ],
+  "scoring_method": "spearman",
+  "primary_score": "task_spearman",
+  "spearman_range": [
+    -1,
+    1
+  ],
+  "spearman_tie_method": "average",
+  "spearman_undefined_policy": "task_error",
+  "spearman_aggregation": "Pearson correlation of true average ranks and submitted positions, separately within each group, then fixed group-weighted sum; do not pool groups or clip negative groups.",
+  "secondary_scoring_method": "equal_pair_inversion_random95",
+  "ranking_reward_range": [
+    0,
+    1
+  ],
+  "random_quantile": 0.95,
+  "unknown_uncertainty_policy": "use_reported_value",
+  "pair_rule": "For each unordered pair, count it iff abs(mu_i-mu_j)>k*s_ij+epsilon for known uncertainty, or >epsilon for unknown uncertainty. Each counted pair has equal weight within its group. Unknown uncertainty is not zero uncertainty.",
+  "ranking_reward_formula": "Validate finite 0.5<=q<1 and 0<=A<=1; if A==1 return 1; else if A<=q+1e-12 return 0; otherwise clamp((A-q)/(1-q),0,1). A is the fixed weighted sum of group pair accuracies; q is the whole-task random 95th percentile.",
+  "format_errors": {
+    "whole_task": "Core=-1; auxiliary=0 for malformed/oversized/non-UTF8 JSON, duplicate object keys, wrong/missing identity, invalid top-level/rankings, unknown or unassignable group.",
+    "known_group": "Missing/duplicate known group or invalid item permutation: raw group Spearman=null, aggregation value=-1, group pair accuracy=0; preserve group weight and apply auxiliary threshold after aggregation."
+  },
+  "max_submission_bytes": 1048576,
+  "minimum_conditions_per_group": 2,
+  "time_limit_sec": 28800,
+  "solver_resources": {
+    "cpus": 8,
+    "memory_mb": 16384,
+    "gpus": 1,
+    "gpu_types": [
+      "H20"
+    ]
+  },
+  "submission_path": "/logs/artifacts/rankings.json",
+  "report_path": "/logs/artifacts/report.md",
+  "sources_path": "/logs/artifacts/sources.json",
+  "analysis_path": "/logs/artifacts/analysis/",
+  "forbidden_result_scope": "Experimental bending stress-strain, flexural failure, toughness or equivalent mechanical rankings for Ag-Cu-Te-Se-S chalcogenide alloy composition and vacancy series are same-type results and prohibited, including related composition or test-protocol variants. General solid mechanics and independent chemical or crystallographic basic facts outside these result panels are allowed."
+}
+```
+<!-- /public-contract -->

tasks/agcute-bending-endpoints-ranking/solution/solve.sh

@@ -0,0 +1,39 @@
+#!/bin/sh
+set -eu
+mkdir -p /logs/artifacts
+cat > /logs/artifacts/rankings.json <<'PROJECTAA_ORACLE'
+{
+  "task_id": "agcute-bending-endpoints-ranking",
+  "task_version": "1.0.0",
+  "rankings": [
+    {
+      "group_id": "max_stress",
+      "order": [
+        "exp_3fed9f8f07c9",
+        "exp_aaa9e2c6f5b9",
+        "exp_3c3c8522da9d",
+        "exp_fc4984656462",
+        "exp_0b361eedc092",
+        "exp_0a801637748b",
+        "exp_65960918c5e1",
+        "exp_39fd1eed6b1c",
+        "exp_365507fb236c"
+      ]
+    },
+    {
+      "group_id": "strain_at_max",
+      "order": [
+        "exp_69a2a5342ec6",
+        "exp_d53afffea8ad",
+        "exp_eb05cc2a9dc8",
+        "exp_4369fc61d66d",
+        "exp_54b8fc8b3571",
+        "exp_2d9d2ad06f6e",
+        "exp_8cfb0d14a259",
+        "exp_3aa8fedf978b",
+        "exp_7b741589c10a"
+      ]
+    }
+  ]
+}
+PROJECTAA_ORACLE

tasks/agcute-bending-endpoints-ranking/task.toml

@@ -0,0 +1,27 @@
+schema_version = "1.2"
+[metadata]
+task_id = "agcute-bending-endpoints-ranking"
+task_version = "1.0.0"
+pipeline_version = "19"
+scientist_prompt_version = "projectaa-scientist-en-v7-ranking"
+[agent]
+user = "solver"
+timeout_sec = 28800.0
+[environment]
+build_timeout_sec = 900.0
+cpus = 8
+memory_mb = 16384
+storage_mb = 20480
+gpus = 1
+gpu_types = ["H20"]
+allow_internet = true
+[verifier]
+timeout_sec = 120.0
+environment_mode = "separate"
+[verifier.environment]
+build_timeout_sec = 600.0
+cpus = 1
+memory_mb = 1024
+storage_mb = 2048
+gpus = 0
+allow_internet = false

tasks/agcute-bending-endpoints-ranking/tests/Dockerfile

@@ -0,0 +1,5 @@
+FROM python@sha256:78387bc3881b8273120a12ebe6c1ab22b018ccc2c9adf565ae1ac9b536e184ea
+ENV PYTHONDONTWRITEBYTECODE=1 PYTHONUNBUFFERED=1
+COPY . /tests/
+RUN chmod +x /tests/test.sh
+WORKDIR /tests

tasks/agcute-bending-endpoints-ranking/tests/answers.json

@@ -0,0 +1,924 @@
+{
+  "task_id": "agcute-bending-endpoints-ranking",
+  "task_version": "1.0.0",
+  "groups": [
+    {
+      "group_id": "max_stress",
+      "items": [
+        {
+          "item_id": "exp_65960918c5e1",
+          "mean": 225.883,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 70960, strain column 7, stress column 8; maximum over 81959 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        },
+        {
+          "item_id": "exp_0b361eedc092",
+          "mean": 203.947,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 112839, strain column 13, stress column 14; maximum over 186708 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        },
+        {
+          "item_id": "exp_0a801637748b",
+          "mean": 208.152,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 119513, strain column 16, stress column 17; maximum over 141872 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        },
+        {
+          "item_id": "exp_39fd1eed6b1c",
+          "mean": 234.551,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 7608, strain column 22, stress column 23; maximum over 7607 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        },
+        {
+          "item_id": "exp_365507fb236c",
+          "mean": 255.958,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 118267, strain column 25, stress column 26; maximum over 156878 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        },
+        {
+          "item_id": "exp_3c3c8522da9d",
+          "mean": 184.302,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 116064, strain column 10, stress column 11; maximum over 126540 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        },
+        {
+          "item_id": "exp_fc4984656462",
+          "mean": 197.459,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 64453, strain column 19, stress column 20; maximum over 64459 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        },
+        {
+          "item_id": "exp_3fed9f8f07c9",
+          "mean": 61.4139,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 5162, strain column 1, stress column 2; maximum over 5162 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        },
+        {
+          "item_id": "exp_aaa9e2c6f5b9",
+          "mean": 148.722,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 13882, strain column 4, stress column 5; maximum over 13881 paired curve points",
+          "aggregation": "Maximum of all finite paired stress samples in the original curve; no interpolation."
+        }
+      ],
+      "pairs": [
+        {
+          "item_i": "exp_65960918c5e1",
+          "item_j": "exp_0b361eedc092",
+          "mean_difference": 21.936000000000007,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_65960918c5e1",
+          "item_j": "exp_0a801637748b",
+          "mean_difference": 17.731000000000023,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_65960918c5e1",
+          "item_j": "exp_39fd1eed6b1c",
+          "mean_difference": -8.667999999999978,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_65960918c5e1",
+          "item_j": "exp_365507fb236c",
+          "mean_difference": -30.07499999999999,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_65960918c5e1",
+          "item_j": "exp_3c3c8522da9d",
+          "mean_difference": 41.58100000000002,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_65960918c5e1",
+          "item_j": "exp_fc4984656462",
+          "mean_difference": 28.424000000000007,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_65960918c5e1",
+          "item_j": "exp_3fed9f8f07c9",
+          "mean_difference": 164.46910000000003,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_65960918c5e1",
+          "item_j": "exp_aaa9e2c6f5b9",
+          "mean_difference": 77.161,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0b361eedc092",
+          "item_j": "exp_0a801637748b",
+          "mean_difference": -4.204999999999984,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0b361eedc092",
+          "item_j": "exp_39fd1eed6b1c",
+          "mean_difference": -30.603999999999985,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0b361eedc092",
+          "item_j": "exp_365507fb236c",
+          "mean_difference": -52.010999999999996,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0b361eedc092",
+          "item_j": "exp_3c3c8522da9d",
+          "mean_difference": 19.64500000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0b361eedc092",
+          "item_j": "exp_fc4984656462",
+          "mean_difference": 6.4879999999999995,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0b361eedc092",
+          "item_j": "exp_3fed9f8f07c9",
+          "mean_difference": 142.5331,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0b361eedc092",
+          "item_j": "exp_aaa9e2c6f5b9",
+          "mean_difference": 55.224999999999994,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0a801637748b",
+          "item_j": "exp_39fd1eed6b1c",
+          "mean_difference": -26.399,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0a801637748b",
+          "item_j": "exp_365507fb236c",
+          "mean_difference": -47.80600000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0a801637748b",
+          "item_j": "exp_3c3c8522da9d",
+          "mean_difference": 23.849999999999994,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0a801637748b",
+          "item_j": "exp_fc4984656462",
+          "mean_difference": 10.692999999999984,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0a801637748b",
+          "item_j": "exp_3fed9f8f07c9",
+          "mean_difference": 146.73809999999997,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_0a801637748b",
+          "item_j": "exp_aaa9e2c6f5b9",
+          "mean_difference": 59.42999999999998,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_39fd1eed6b1c",
+          "item_j": "exp_365507fb236c",
+          "mean_difference": -21.40700000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_39fd1eed6b1c",
+          "item_j": "exp_3c3c8522da9d",
+          "mean_difference": 50.248999999999995,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_39fd1eed6b1c",
+          "item_j": "exp_fc4984656462",
+          "mean_difference": 37.091999999999985,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_39fd1eed6b1c",
+          "item_j": "exp_3fed9f8f07c9",
+          "mean_difference": 173.13709999999998,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_39fd1eed6b1c",
+          "item_j": "exp_aaa9e2c6f5b9",
+          "mean_difference": 85.82899999999998,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_365507fb236c",
+          "item_j": "exp_3c3c8522da9d",
+          "mean_difference": 71.656,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_365507fb236c",
+          "item_j": "exp_fc4984656462",
+          "mean_difference": 58.498999999999995,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_365507fb236c",
+          "item_j": "exp_3fed9f8f07c9",
+          "mean_difference": 194.54410000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_365507fb236c",
+          "item_j": "exp_aaa9e2c6f5b9",
+          "mean_difference": 107.23599999999999,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3c3c8522da9d",
+          "item_j": "exp_fc4984656462",
+          "mean_difference": -13.15700000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3c3c8522da9d",
+          "item_j": "exp_3fed9f8f07c9",
+          "mean_difference": 122.8881,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3c3c8522da9d",
+          "item_j": "exp_aaa9e2c6f5b9",
+          "mean_difference": 35.579999999999984,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_fc4984656462",
+          "item_j": "exp_3fed9f8f07c9",
+          "mean_difference": 136.0451,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_fc4984656462",
+          "item_j": "exp_aaa9e2c6f5b9",
+          "mean_difference": 48.736999999999995,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3fed9f8f07c9",
+          "item_j": "exp_aaa9e2c6f5b9",
+          "mean_difference": -87.30810000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        }
+      ],
+      "comparable_pairs": 36,
+      "quality_status": "passed"
+    },
+    {
+      "group_id": "strain_at_max",
+      "items": [
+        {
+          "item_id": "exp_54b8fc8b3571",
+          "mean": 3.58553,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 70960, strain column 7, stress column 8; maximum over 81959 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        },
+        {
+          "item_id": "exp_3aa8fedf978b",
+          "mean": 6.20628,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 118267, strain column 25, stress column 26; maximum over 156878 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        },
+        {
+          "item_id": "exp_2d9d2ad06f6e",
+          "mean": 5.59208,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 112839, strain column 13, stress column 14; maximum over 186708 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        },
+        {
+          "item_id": "exp_8cfb0d14a259",
+          "mean": 6.15589,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 119513, strain column 16, stress column 17; maximum over 141872 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        },
+        {
+          "item_id": "exp_d53afffea8ad",
+          "mean": 0.67426,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 13882, strain column 4, stress column 5; maximum over 13881 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        },
+        {
+          "item_id": "exp_69a2a5342ec6",
+          "mean": 0.26561,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 5162, strain column 1, stress column 2; maximum over 5162 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        },
+        {
+          "item_id": "exp_eb05cc2a9dc8",
+          "mean": 1.57783,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 7608, strain column 22, stress column 23; maximum over 7607 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        },
+        {
+          "item_id": "exp_7b741589c10a",
+          "mean": 6.31575,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 116064, strain column 10, stress column 11; maximum over 126540 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        },
+        {
+          "item_id": "exp_4369fc61d66d",
+          "mean": 3.19442,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Unknown independent uncertainty; use reported values without inventing zero uncertainty."
+          ],
+          "source_location": "Original-SourceData.xlsx 'Figure 1b' row 64453, strain column 19, stress column 20; maximum over 64459 paired curve points",
+          "aggregation": "Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation."
+        }
+      ],
+      "pairs": [
+        {
+          "item_i": "exp_54b8fc8b3571",
+          "item_j": "exp_3aa8fedf978b",
+          "mean_difference": -2.6207499999999997,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_54b8fc8b3571",
+          "item_j": "exp_2d9d2ad06f6e",
+          "mean_difference": -2.0065500000000003,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_54b8fc8b3571",
+          "item_j": "exp_8cfb0d14a259",
+          "mean_difference": -2.5703600000000004,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_54b8fc8b3571",
+          "item_j": "exp_d53afffea8ad",
+          "mean_difference": 2.91127,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_54b8fc8b3571",
+          "item_j": "exp_69a2a5342ec6",
+          "mean_difference": 3.3199199999999998,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_54b8fc8b3571",
+          "item_j": "exp_eb05cc2a9dc8",
+          "mean_difference": 2.0077,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_54b8fc8b3571",
+          "item_j": "exp_7b741589c10a",
+          "mean_difference": -2.7302200000000005,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_54b8fc8b3571",
+          "item_j": "exp_4369fc61d66d",
+          "mean_difference": 0.39110999999999985,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3aa8fedf978b",
+          "item_j": "exp_2d9d2ad06f6e",
+          "mean_difference": 0.6141999999999994,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3aa8fedf978b",
+          "item_j": "exp_8cfb0d14a259",
+          "mean_difference": 0.05038999999999927,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3aa8fedf978b",
+          "item_j": "exp_d53afffea8ad",
+          "mean_difference": 5.532019999999999,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3aa8fedf978b",
+          "item_j": "exp_69a2a5342ec6",
+          "mean_difference": 5.94067,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3aa8fedf978b",
+          "item_j": "exp_eb05cc2a9dc8",
+          "mean_difference": 4.628449999999999,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3aa8fedf978b",
+          "item_j": "exp_7b741589c10a",
+          "mean_difference": -0.10947000000000084,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_3aa8fedf978b",
+          "item_j": "exp_4369fc61d66d",
+          "mean_difference": 3.0118599999999995,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_2d9d2ad06f6e",
+          "item_j": "exp_8cfb0d14a259",
+          "mean_difference": -0.5638100000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_2d9d2ad06f6e",
+          "item_j": "exp_d53afffea8ad",
+          "mean_difference": 4.91782,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_2d9d2ad06f6e",
+          "item_j": "exp_69a2a5342ec6",
+          "mean_difference": 5.3264700000000005,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_2d9d2ad06f6e",
+          "item_j": "exp_eb05cc2a9dc8",
+          "mean_difference": 4.0142500000000005,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_2d9d2ad06f6e",
+          "item_j": "exp_7b741589c10a",
+          "mean_difference": -0.7236700000000003,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_2d9d2ad06f6e",
+          "item_j": "exp_4369fc61d66d",
+          "mean_difference": 2.39766,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8cfb0d14a259",
+          "item_j": "exp_d53afffea8ad",
+          "mean_difference": 5.48163,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8cfb0d14a259",
+          "item_j": "exp_69a2a5342ec6",
+          "mean_difference": 5.890280000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8cfb0d14a259",
+          "item_j": "exp_eb05cc2a9dc8",
+          "mean_difference": 4.578060000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8cfb0d14a259",
+          "item_j": "exp_7b741589c10a",
+          "mean_difference": -0.1598600000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8cfb0d14a259",
+          "item_j": "exp_4369fc61d66d",
+          "mean_difference": 2.9614700000000003,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_d53afffea8ad",
+          "item_j": "exp_69a2a5342ec6",
+          "mean_difference": 0.40864999999999996,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_d53afffea8ad",
+          "item_j": "exp_eb05cc2a9dc8",
+          "mean_difference": -0.9035700000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_d53afffea8ad",
+          "item_j": "exp_7b741589c10a",
+          "mean_difference": -5.64149,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_d53afffea8ad",
+          "item_j": "exp_4369fc61d66d",
+          "mean_difference": -2.52016,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_69a2a5342ec6",
+          "item_j": "exp_eb05cc2a9dc8",
+          "mean_difference": -1.31222,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_69a2a5342ec6",
+          "item_j": "exp_7b741589c10a",
+          "mean_difference": -6.050140000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_69a2a5342ec6",
+          "item_j": "exp_4369fc61d66d",
+          "mean_difference": -2.92881,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_eb05cc2a9dc8",
+          "item_j": "exp_7b741589c10a",
+          "mean_difference": -4.737920000000001,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_eb05cc2a9dc8",
+          "item_j": "exp_4369fc61d66d",
+          "mean_difference": -1.61659,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_7b741589c10a",
+          "item_j": "exp_4369fc61d66d",
+          "mean_difference": 3.1213300000000004,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        }
+      ],
+      "comparable_pairs": 36,
+      "quality_status": "passed"
+    }
+  ]
+}

tasks/agcute-bending-endpoints-ranking/tests/contract.json

@@ -0,0 +1,98 @@
+{
+  "task_id": "agcute-bending-endpoints-ranking",
+  "task_version": "1.0.0",
+  "groups": [
+    {
+      "group_id": "max_stress",
+      "label": "Alloy composition series: bending stress",
+      "item_ids": [
+        "exp_65960918c5e1",
+        "exp_0b361eedc092",
+        "exp_0a801637748b",
+        "exp_39fd1eed6b1c",
+        "exp_365507fb236c",
+        "exp_3c3c8522da9d",
+        "exp_fc4984656462",
+        "exp_3fed9f8f07c9",
+        "exp_aaa9e2c6f5b9"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.5,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "Maximum bending stress",
+        "unit": "MPa",
+        "definition_en": "Largest recorded engineering bending stress in each curve."
+      },
+      "uncertainty_definition": "Reliable independent uncertainty is not available for this panel. Unequal reported values are compared directly; uncertainty is not assumed to be zero."
+    },
+    {
+      "group_id": "strain_at_max",
+      "label": "Alloy composition series: peak-stress strain",
+      "item_ids": [
+        "exp_54b8fc8b3571",
+        "exp_3aa8fedf978b",
+        "exp_2d9d2ad06f6e",
+        "exp_8cfb0d14a259",
+        "exp_d53afffea8ad",
+        "exp_69a2a5342ec6",
+        "exp_eb05cc2a9dc8",
+        "exp_7b741589c10a",
+        "exp_4369fc61d66d"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.5,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "Engineering strain at maximum bending stress",
+        "unit": "%",
+        "definition_en": "Engineering strain at the same sampled point as the largest bending stress, using the earliest occurrence if maxima repeat."
+      },
+      "uncertainty_definition": "Reliable independent uncertainty is not available for this panel. Unequal reported values are compared directly; uncertainty is not assumed to be zero."
+    }
+  ],
+  "scoring_method": "spearman",
+  "primary_score": "task_spearman",
+  "spearman_range": [
+    -1,
+    1
+  ],
+  "spearman_tie_method": "average",
+  "spearman_undefined_policy": "task_error",
+  "spearman_aggregation": "Pearson correlation of true average ranks and submitted positions, separately within each group, then fixed group-weighted sum; do not pool groups or clip negative groups.",
+  "secondary_scoring_method": "equal_pair_inversion_random95",
+  "ranking_reward_range": [
+    0,
+    1
+  ],
+  "random_quantile": 0.95,
+  "unknown_uncertainty_policy": "use_reported_value",
+  "pair_rule": "For each unordered pair, count it iff abs(mu_i-mu_j)>k*s_ij+epsilon for known uncertainty, or >epsilon for unknown uncertainty. Each counted pair has equal weight within its group. Unknown uncertainty is not zero uncertainty.",
+  "ranking_reward_formula": "Validate finite 0.5<=q<1 and 0<=A<=1; if A==1 return 1; else if A<=q+1e-12 return 0; otherwise clamp((A-q)/(1-q),0,1). A is the fixed weighted sum of group pair accuracies; q is the whole-task random 95th percentile.",
+  "format_errors": {
+    "whole_task": "Core=-1; auxiliary=0 for malformed/oversized/non-UTF8 JSON, duplicate object keys, wrong/missing identity, invalid top-level/rankings, unknown or unassignable group.",
+    "known_group": "Missing/duplicate known group or invalid item permutation: raw group Spearman=null, aggregation value=-1, group pair accuracy=0; preserve group weight and apply auxiliary threshold after aggregation."
+  },
+  "max_submission_bytes": 1048576,
+  "minimum_conditions_per_group": 2,
+  "time_limit_sec": 28800,
+  "solver_resources": {
+    "cpus": 8,
+    "memory_mb": 16384,
+    "gpus": 1,
+    "gpu_types": [
+      "H20"
+    ]
+  },
+  "submission_path": "/logs/artifacts/rankings.json",
+  "report_path": "/logs/artifacts/report.md",
+  "sources_path": "/logs/artifacts/sources.json",
+  "analysis_path": "/logs/artifacts/analysis/",
+  "forbidden_result_scope": "Experimental bending stress-strain, flexural failure, toughness or equivalent mechanical rankings for Ag-Cu-Te-Se-S chalcogenide alloy composition and vacancy series are same-type results and prohibited, including related composition or test-protocol variants. General solid mechanics and independent chemical or crystallographic basic facts outside these result panels are allowed."
+}

tasks/agcute-bending-endpoints-ranking/tests/grade.py

@@ -0,0 +1,174 @@
+"""Pipeline v18 Spearman core and preserved equal-pair auxiliary grader."""
+import argparse
+import copy
+import hashlib
+import json
+import math
+from pathlib import Path
+
+import original_grade as old
+
+TaskError = old.TaskError
+
+
+def bounded(value):
+    if not math.isfinite(value) or not -1 - 1e-12 <= value <= 1 + 1e-12:
+        raise TaskError('Spearman outside its finite range')
+    return min(1.0, max(-1.0, value))
+
+
+def average_ranks(values):
+    if len(values) < 2 or any(isinstance(x, bool) or not isinstance(x, (int, float))
+                              or not math.isfinite(x) for x in values):
+        raise TaskError('Invalid measurements')
+    ordered = sorted(range(len(values)), key=values.__getitem__)
+    ranks = [0.0] * len(values)
+    start = 0
+    while start < len(values):
+        stop = start + 1
+        while stop < len(values) and values[ordered[stop]] == values[ordered[start]]:
+            stop += 1
+        mean_rank = (start + 1 + stop) / 2
+        for index in ordered[start:stop]:
+            ranks[index] = mean_rank
+        start = stop
+    return ranks
+
+
+def spearman(values, predicted_positions):
+    ranks = average_ranks(values)
+    n = len(values)
+    if sorted(predicted_positions) != list(range(n)):
+        raise ValueError('Not a complete permutation')
+    predicted_ranks = [p + 1 for p in predicted_positions]
+    true_mean = math.fsum(ranks) / n
+    predicted_mean = math.fsum(predicted_ranks) / n
+    x = [r - true_mean for r in ranks]
+    y = [p - predicted_mean for p in predicted_ranks]
+    denominator = math.sqrt(math.fsum(r * r for r in x) * math.fsum(p * p for p in y))
+    if denominator <= 0:
+        raise TaskError('Undefined Spearman for constant results')
+    return bounded(math.fsum(r * p for r, p in zip(x, y)) / denominator)
+
+
+def promote_contract(contract):
+    value = copy.deepcopy(contract)
+    value.update(scoring_method='spearman', primary_score='task_spearman',
+                 spearman_range=[-1, 1], spearman_tie_method='average',
+                 spearman_undefined_policy='task_error',
+                 secondary_scoring_method='equal_pair_inversion_random95',
+                 ranking_reward_range=[0, 1])
+    return value
+
+
+def auxiliary_view(contract):
+    expected = {'scoring_method': 'spearman', 'primary_score': 'task_spearman',
+                'spearman_range': [-1, 1], 'spearman_tie_method': 'average',
+                'spearman_undefined_policy': 'task_error',
+                'secondary_scoring_method': 'equal_pair_inversion_random95',
+                'ranking_reward_range': [0, 1]}
+    if any(contract.get(k) != v for k, v in expected.items()):
+        raise TaskError('Invalid dual-score contract')
+    value = copy.deepcopy(contract)
+    value['scoring_method'] = value.pop('secondary_scoring_method')
+    for key in ('primary_score', 'spearman_range', 'spearman_tie_method',
+                'spearman_undefined_policy', 'ranking_reward_range'):
+        value.pop(key)
+    return value
+
+
+def scoring_bindings(contract, answers):
+    return {'contract_sha256': old.canonical_sha(contract),
+            'answers_sha256': old.canonical_sha(answers),
+            'grader_sha256': hashlib.sha256(Path(__file__).read_bytes()).hexdigest(),
+            'auxiliary_grader_sha256': hashlib.sha256(Path(old.__file__).read_bytes()).hexdigest()}
+
+
+def validate(contract, answers, baseline):
+    auxiliary = auxiliary_view(contract)
+    old.validate_task(auxiliary, answers)
+    if baseline.get('bindings') != scoring_bindings(contract, answers):
+        raise TaskError('Dual-score baseline hash mismatch')
+    old_baseline = {'q': baseline.get('q'), 'bindings': old.scoring_bindings(auxiliary, answers)}
+    old.validate_baseline(auxiliary, answers, old_baseline)
+    hidden = {g['group_id']: {i['item_id']: i['mean'] for i in g['items']} for g in answers['groups']}
+    for g in contract['groups']:
+        spearman([hidden[g['group_id']][i] for i in g['item_ids']], list(range(len(g['item_ids']))))
+    return auxiliary, old_baseline, hidden
+
+
+def add_core(contract, report, hidden):
+    report['scoring_method'] = 'spearman'
+    report['primary_score'] = 'task_spearman'
+    report['secondary_scoring_method'] = 'equal_pair_inversion_random95'
+    report['official_reward'] = None
+    report['official_ranking_reward'] = None
+    if not report['groups']:
+        report['task_spearman'] = -1.0
+        report['core_penalty_reason'] = 'whole_task_format_error'
+        report['groups'] = [{'group_id': g['group_id'], 'group_weight': g['group_weight'],
+                             'group_spearman_g': None, 'spearman_for_aggregation': -1.0,
+                             'group_pair_accuracy': 0.0, 'status': 'whole_task_format_error'}
+                            for g in contract['groups']]
+        return report
+    for g in report['groups']:
+        if g['status'] != 'valid':
+            rho = None
+        else:
+            item_ids = next(c['item_ids'] for c in contract['groups'] if c['group_id'] == g['group_id'])
+            positions = {item_id: i for i, item_id in enumerate(g['order'])}
+            rho = spearman([hidden[g['group_id']][i] for i in item_ids], [positions[i] for i in item_ids])
+        g['group_spearman_g'] = rho
+        g['spearman_for_aggregation'] = -1.0 if rho is None else rho
+    report['task_spearman'] = bounded(math.fsum(g['group_weight'] * g['spearman_for_aggregation']
+                                               for g in report['groups']))
+    return report
+
+
+def grade_submission(contract, answers, submission, baseline):
+    auxiliary, old_baseline, hidden = validate(contract, answers, baseline)
+    report = old.grade_submission(auxiliary, answers, submission, old_baseline)
+    report['random_q95'] = baseline['q']
+    return add_core(contract, report, hidden)
+
+
+def grade_file(contract, answers, path, baseline):
+    auxiliary, old_baseline, hidden = validate(contract, answers, baseline)
+    return add_core(contract, old.grade_file(auxiliary, answers, path, old_baseline), hidden)
+
+
+def official_scores(report, compliance, execution, memory=False):
+    if compliance == 'cheating':
+        return -1.0, 0.0
+    if compliance != 'compliant' or execution != 'completed':
+        return None, None
+    return report['task_spearman'], report['ranking_reward']
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--interface', choices=['txt', 'json'], default='txt')
+    parser.add_argument('--root', type=Path, default=Path('/tests'))
+    parser.add_argument('--submission', type=Path, default=Path('/logs/artifacts/rankings.json'))
+    parser.add_argument('--output', type=Path, default=Path('/logs/verifier'))
+    args = parser.parse_args()
+    args.output.mkdir(parents=True, exist_ok=True)
+    for name in ('reward.txt', 'reward.json', 'report.json', 'evaluator_error.json'):
+        (args.output / name).unlink(missing_ok=True)
+    try:
+        contract, answers, baseline = [old.load_json((args.root / f'{name}.json').read_bytes())
+                                      for name in ('contract', 'answers', 'random_baseline')]
+        report = grade_file(contract, answers, args.submission, baseline)
+    except (TaskError, old.SubmissionError, OSError, ValueError, KeyError, TypeError) as error:
+        old.dump(args.output / 'evaluator_error.json', {'status': 'task_error', 'error': str(error)})
+        raise SystemExit(2) from error
+    old.dump(args.output / 'report.json', report)
+    if args.interface == 'txt':
+        (args.output / 'reward.txt').write_text(str(report['task_spearman']) + '\n')
+    else:
+        old.dump(args.output / 'reward.json', {'reward': report['task_spearman'],
+                                              'ranking_reward': report['ranking_reward']})
+
+
+if __name__ == '__main__':
+    main()

tasks/agcute-bending-endpoints-ranking/tests/measurements.csv

@@ -0,0 +1,19 @@
+group_id,item_id,mean,n,uncertainty_status,uncertainty_scale,uncertainty_basis,quality_status,warnings,source_location,aggregation
+max_stress,exp_65960918c5e1,225.883,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 70960, strain column 7, stress column 8; maximum over 81959 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+max_stress,exp_0b361eedc092,203.947,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 112839, strain column 13, stress column 14; maximum over 186708 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+max_stress,exp_0a801637748b,208.152,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 119513, strain column 16, stress column 17; maximum over 141872 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+max_stress,exp_39fd1eed6b1c,234.551,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 7608, strain column 22, stress column 23; maximum over 7607 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+max_stress,exp_365507fb236c,255.958,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 118267, strain column 25, stress column 26; maximum over 156878 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+max_stress,exp_3c3c8522da9d,184.302,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 116064, strain column 10, stress column 11; maximum over 126540 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+max_stress,exp_fc4984656462,197.459,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 64453, strain column 19, stress column 20; maximum over 64459 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+max_stress,exp_3fed9f8f07c9,61.4139,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 5162, strain column 1, stress column 2; maximum over 5162 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+max_stress,exp_aaa9e2c6f5b9,148.722,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 13882, strain column 4, stress column 5; maximum over 13881 paired curve points",Maximum of all finite paired stress samples in the original curve; no interpolation.
+strain_at_max,exp_54b8fc8b3571,3.58553,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 70960, strain column 7, stress column 8; maximum over 81959 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.
+strain_at_max,exp_3aa8fedf978b,6.20628,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 118267, strain column 25, stress column 26; maximum over 156878 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.
+strain_at_max,exp_2d9d2ad06f6e,5.59208,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 112839, strain column 13, stress column 14; maximum over 186708 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.
+strain_at_max,exp_8cfb0d14a259,6.15589,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 119513, strain column 16, stress column 17; maximum over 141872 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.
+strain_at_max,exp_d53afffea8ad,0.67426,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 13882, strain column 4, stress column 5; maximum over 13881 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.
+strain_at_max,exp_69a2a5342ec6,0.26561,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 5162, strain column 1, stress column 2; maximum over 5162 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.
+strain_at_max,exp_eb05cc2a9dc8,1.57783,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 7608, strain column 22, stress column 23; maximum over 7607 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.
+strain_at_max,exp_7b741589c10a,6.31575,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 116064, strain column 10, stress column 11; maximum over 126540 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.
+strain_at_max,exp_4369fc61d66d,3.19442,,unknown,,No reliable independent difference uncertainty is supplied for this selected endpoint. Serial curve points are not independent sample replicates.,retained_with_flag,"[""Unknown independent uncertainty; use reported values without inventing zero uncertainty.""]","Original-SourceData.xlsx 'Figure 1b' row 64453, strain column 19, stress column 20; maximum over 64459 paired curve points",Engineering strain at the first occurrence of the original curve's maximum stress; no interpolation.

tasks/agcute-bending-endpoints-ranking/tests/original_grade.py

@@ -0,0 +1,423 @@
+"""Pipeline v16: equal pairs within groups; one task-level random-95 threshold."""
+
+import argparse
+import errno
+import hashlib
+import itertools
+import json
+import math
+import os
+from pathlib import Path
+import stat
+
+
+class TaskError(ValueError):
+    pass
+
+
+class SubmissionError(ValueError):
+    pass
+
+
+def canonical_sha(value):
+    return hashlib.sha256(json.dumps(value, sort_keys=True, separators=(",", ":"),
+                                     ensure_ascii=False, allow_nan=False).encode()).hexdigest()
+
+
+def scoring_bindings(contract, answers):
+    return {
+        "contract_sha256": canonical_sha(contract),
+        "answers_sha256": canonical_sha(answers),
+        "relations_sha256": canonical_sha([{"group_id": g["group_id"], "pairs": g["pairs"]}
+                                           for g in answers["groups"]]),
+        "weights_sha256": canonical_sha([{k: g[k] for k in ("group_id", "group_weight")}
+                                         for g in contract["groups"]]),
+        "grader_sha256": hashlib.sha256(Path(__file__).read_bytes()).hexdigest(),
+    }
+
+
+def validate_baseline(contract, answers, baseline):
+    if not isinstance(baseline, dict):
+        raise TaskError("random baseline must be an object")
+    q = finite(baseline.get("q"), "random_q95", 0.5)
+    if q >= 1:
+        raise TaskError("random_q95 must be below 1")
+    if contract.get("random_quantile") != 0.95:
+        raise TaskError("invalid random quantile")
+    if baseline.get("bindings") != scoring_bindings(contract, answers):
+        raise TaskError("random baseline hash mismatch")
+    return q
+
+
+def reward_from_accuracy(accuracy, q):
+    accuracy = finite(accuracy, "accuracy", 0)
+    q = finite(q, "random_q95", 0.5)
+    if accuracy > 1 or q >= 1:
+        raise TaskError("invalid accuracy or random threshold")
+    if accuracy == 1:
+        return 1.0
+    if accuracy <= q + 1e-12:
+        return 0.0
+    return min(1.0, max(0.0, (accuracy - q) / (1 - q)))
+
+
+def aggregate_accuracy(details):
+    # Preserve exact endpoints despite representational roundoff in weights.
+    if all(g["group_pair_accuracy"] == 1 for g in details):
+        return 1.0
+    return math.fsum(g["group_pair_accuracy"] * g["group_weight"] for g in details)
+
+
+def unique_object(pairs):
+    result = {}
+    for key, value in pairs:
+        if key in result:
+            raise SubmissionError("duplicate_json_key")
+        result[key] = value
+    return result
+
+
+def load_json(raw):
+    def reject_constant(value):
+        raise SubmissionError("nonfinite_json_constant")
+    return json.loads(raw.decode("utf-8"), object_pairs_hook=unique_object,
+                      parse_constant=reject_constant)
+
+
+def finite(value, name, minimum=None):
+    if isinstance(value, bool) or not isinstance(value, (int, float)):
+        raise TaskError(name + " is not numeric")
+    if not math.isfinite(value) or (minimum is not None and value < minimum):
+        raise TaskError(name + " is out of range")
+    return float(value)
+
+
+def difference_se(sem_i, sem_j, method):
+    if method == "sem_sum_upper_bound":
+        return sem_i + sem_j
+    if method == "independent_sem_quadrature":
+        return math.hypot(sem_i, sem_j)
+    raise TaskError("unsupported difference standard error method")
+
+
+def validate_task(contract, answers):
+    if not isinstance(contract, dict) or not isinstance(answers, dict):
+        raise TaskError("contract and answers must be objects")
+    if contract.get("scoring_method") != "equal_pair_inversion_random95":
+        raise TaskError("unsupported scoring method")
+    if contract.get("unknown_uncertainty_policy") != "use_reported_value":
+        raise TaskError("unsupported unknown-uncertainty policy")
+    for key in ("task_id", "task_version"):
+        if not isinstance(contract.get(key), str) or answers.get(key) != contract[key]:
+            raise TaskError("invalid task identity")
+    limit = contract.get("max_submission_bytes")
+    if type(limit) is not int or not 0 < limit <= 1048576:
+        raise TaskError("invalid submission size limit")
+    minimum_conditions = contract.get("minimum_conditions_per_group", 2)
+    if type(minimum_conditions) is not int or minimum_conditions < 2:
+        raise TaskError("invalid minimum conditions")
+    groups = contract.get("groups")
+    hidden_groups = answers.get("groups")
+    if not isinstance(groups, list) or not groups or not isinstance(hidden_groups, list):
+        raise TaskError("missing groups")
+    if len(groups) != len(hidden_groups):
+        raise TaskError("hidden group count mismatch")
+    hidden_map = {}
+    for hidden in hidden_groups:
+        if not isinstance(hidden, dict) or not isinstance(hidden.get("group_id"), str):
+            raise TaskError("invalid hidden group")
+        if hidden["group_id"] in hidden_map:
+            raise TaskError("duplicate hidden group")
+        hidden_map[hidden["group_id"]] = hidden
+    validated, group_ids, all_ids = [], set(), set()
+    for group in groups:
+        if not isinstance(group, dict):
+            raise TaskError("invalid contract group")
+        gid = group.get("group_id")
+        if not isinstance(gid, str) or not gid or gid in group_ids:
+            raise TaskError("invalid group id")
+        group_ids.add(gid)
+        ids = group.get("item_ids")
+        if (not isinstance(ids, list) or len(ids) < minimum_conditions
+                or any(not isinstance(item, str) or not item for item in ids)
+                or len(ids) != len(set(ids)) or all_ids.intersection(ids)):
+            raise TaskError("invalid item ids")
+        all_ids.update(ids)
+        if group.get("direction") != "ascending":
+            raise TaskError("only ascending order is supported")
+        group_weight = finite(group.get("group_weight"), "group_weight", 0)
+        if group_weight == 0:
+            raise TaskError("zero group weight")
+        k = finite(group.get("k"), "k", 0)
+        epsilon = finite(group.get("epsilon"), "epsilon", 0)
+        method = group.get("difference_standard_error")
+        if method not in ("sem_sum_upper_bound", "independent_sem_quadrature", "unknown"):
+            raise TaskError("unsupported difference standard error method")
+        if group.get("uncertainty_model") not in ("optional_reported_uncertainty",
+                "reported_sd_with_source_bounds", "independent_replicate_mean", "replicate_sd_upper_envelope", "unknown_replicate_hierarchy"):
+            raise TaskError("unsupported marginal uncertainty model")
+        hidden = hidden_map.get(gid)
+        if hidden is None or hidden.get("quality_status") != "passed":
+            raise TaskError("missing or invalid hidden group")
+        items = hidden.get("items")
+        if not isinstance(items, list) or len(items) != len(ids):
+            raise TaskError("hidden item count mismatch")
+        item_map = {}
+        for item in items:
+            if not isinstance(item, dict) or item.get("item_id") not in ids:
+                raise TaskError("invalid hidden item")
+            item_id = item["item_id"]
+            if item_id in item_map or item.get("quality_status") not in ("retained", "retained_with_flag"):
+                raise TaskError("duplicate or invalid hidden item")
+            mean = finite(item.get("mean"), "mean")
+            n = item.get("n")
+            if n is not None and (type(n) is not int or n < 1):
+                raise TaskError("invalid experimental replicate count")
+            uncertainty_status = item.get("uncertainty_status")
+            if uncertainty_status not in ("known", "unknown"):
+                raise TaskError("invalid uncertainty status")
+            uncertainty_model = group.get("uncertainty_model", "independent_replicate_mean")
+            if uncertainty_model == "reported_sd_with_source_bounds":
+                lo = finite(item.get("mean_lower"), "mean_lower")
+                hi = finite(item.get("mean_upper"), "mean_upper")
+                if lo > hi or not lo - 1e-12 <= mean <= hi + 1e-12:
+                    raise TaskError("inconsistent reported mean interval")
+            if uncertainty_status == "unknown":
+                if item.get("uncertainty_scale") is not None:
+                    raise TaskError("unknown uncertainty must be null")
+                sem = None
+            elif uncertainty_model == "optional_reported_uncertainty":
+                sem = finite(item.get("uncertainty_scale"), "uncertainty_scale", 0)
+                if not item.get("uncertainty_basis"):
+                    raise TaskError("known uncertainty needs a stated source basis")
+            elif uncertainty_model == "reported_sd_with_source_bounds":
+                if n is None or n < 2:
+                    raise TaskError("reported SD needs repeat evidence")
+                lower_sd = finite(item.get("sd_lower"), "sd_lower", 0)
+                upper_sd = finite(item.get("sd_upper"), "sd_upper", 0)
+                if lower_sd > upper_sd:
+                    raise TaskError("inconsistent reported SD interval")
+                sem = finite(item.get("uncertainty_scale"), "uncertainty_scale", 0)
+                expected_scale = upper_sd + (hi - lo) / 2
+                if not math.isclose(sem, expected_scale, rel_tol=1e-12, abs_tol=1e-12):
+                    raise TaskError("inconsistent source-bound uncertainty")
+            elif uncertainty_model in ("independent_replicate_mean", "replicate_sd_upper_envelope"):
+                sem = finite(item.get("uncertainty_scale"), "uncertainty_scale", 0)
+                sd = finite(item.get("sample_sd"), "sample_sd", 0)
+                if n is None or n < 2:
+                    raise TaskError("known SD needs repeat evidence")
+                expected_sem = sd / math.sqrt(n) if uncertainty_model == "independent_replicate_mean" else sd
+                if not math.isclose(sem, expected_sem, rel_tol=1e-12, abs_tol=1e-12):
+                    raise TaskError("inconsistent hidden uncertainty")
+            else:
+                raise TaskError("unsupported marginal uncertainty model")
+            item_map[item_id] = {"mean": mean, "sem": sem, "uncertainty_status": uncertainty_status}
+        stored = hidden.get("pairs")
+        if not isinstance(stored, list) or len(stored) != len(ids) * (len(ids) - 1) // 2:
+            raise TaskError("invalid hidden pair count")
+        stored_map = {}
+        for pair in stored:
+            if not isinstance(pair, dict):
+                raise TaskError("invalid hidden pair")
+            key = (pair.get("item_i"), pair.get("item_j"))
+            if key in stored_map:
+                raise TaskError("duplicate hidden pair")
+            stored_map[key] = pair
+        pairs = []
+        for item_i, item_j in itertools.combinations(ids, 2):
+            left, right = item_map[item_i], item_map[item_j]
+            difference = left["mean"] - right["mean"]
+            known = left["uncertainty_status"] == right["uncertainty_status"] == "known"
+            se = difference_se(left["sem"], right["sem"], method) if known else None
+            tolerance = k * se + epsilon if known else epsilon
+            comparable = int(abs(difference) > tolerance)
+            basis = "uncertainty_adjusted" if known else "reported_value"
+            if not math.isfinite(difference) or not math.isfinite(tolerance) or (known and not math.isfinite(se)):
+                raise TaskError("nonfinite pair arithmetic")
+            pair = stored_map.get((item_i, item_j))
+            if pair is None:
+                raise TaskError("missing hidden pair")
+            for name, expected, minimum in (("mean_difference", difference, None),):
+                actual = finite(pair.get(name), name, minimum)
+                if not math.isclose(actual, expected, rel_tol=1e-12, abs_tol=1e-10):
+                    raise TaskError("inconsistent hidden " + name)
+            if known:
+                if not math.isclose(finite(pair.get("difference_se"), "difference_se", 0), se,
+                                    rel_tol=1e-12, abs_tol=1e-10):
+                    raise TaskError("inconsistent pair uncertainty")
+            elif pair.get("difference_se") is not None:
+                raise TaskError("unknown pair uncertainty must be null")
+            if type(pair.get("comparable")) is not int or pair["comparable"] != comparable:
+                raise TaskError("inconsistent comparable flag")
+            if pair.get("uncertainty_status") != ("known" if known else "unknown"):
+                raise TaskError("inconsistent pair uncertainty status")
+            if pair.get("comparison_basis") != basis:
+                raise TaskError("inconsistent comparison basis")
+            pairs.append({"item_i": item_i, "item_j": item_j,
+                          "mean_difference": difference, "difference_se": se,
+                          "uncertainty_status": "known" if known else "unknown",
+                          "comparison_basis": basis, "comparable": comparable})
+        total = sum(pair["comparable"] for pair in pairs)
+        if total <= 0:
+            raise TaskError("group has no scoreable pair under the declared rule")
+        if type(hidden.get("comparable_pairs")) is not int or hidden["comparable_pairs"] != total:
+            raise TaskError("inconsistent comparable pair count")
+        validated.append({"group_id": gid, "item_ids": ids, "group_weight": group_weight,
+                          "pairs": pairs, "comparable_pairs": total})
+    if set(hidden_map) != group_ids:
+        raise TaskError("unknown hidden group")
+    if not math.isclose(math.fsum(g["group_weight"] for g in validated), 1.0,
+                        rel_tol=0, abs_tol=1e-12):
+        raise TaskError("group weights must sum to one")
+    return validated
+
+
+def zero_report(contract, reason):
+    return {"task_id": contract["task_id"], "task_version": contract["task_version"],
+            "ranking_reward": 0.0, "submission_status": reason, "groups": [],
+            "task_pair_accuracy": 0.0, "aggregation_zero_due_to_format": True,
+            "official_reward": None, "review_status": "not_reviewed", "evaluation_status": "completed"}
+
+
+def grade_submission(contract, answers, submission, baseline):
+    groups = validate_task(contract, answers)
+    q = validate_baseline(contract, answers, baseline)
+    if not isinstance(submission, dict):
+        return zero_report(contract, "invalid_top_level_structure")
+    if any(submission.get(key) != contract[key] for key in ("task_id", "task_version")):
+        return zero_report(contract, "task_identity_mismatch")
+    entries = submission.get("rankings")
+    if not isinstance(entries, list):
+        return zero_report(contract, "rankings_not_array")
+    known = {g["group_id"] for g in groups}
+    by_group = {}
+    for entry in entries:
+        if not isinstance(entry, dict) or not isinstance(entry.get("group_id"), str):
+            return zero_report(contract, "unassignable_group_entry")
+        if entry["group_id"] not in known:
+            return zero_report(contract, "unknown_group_id")
+        by_group.setdefault(entry["group_id"], []).append(entry)
+    details = []
+    for group in groups:
+        gid = group["group_id"]
+        matches = by_group.get(gid, [])
+        item = {"group_id": gid, "group_weight": group["group_weight"],
+                "group_pair_accuracy": 0.0, "status": "missing_group", "comparable_pairs": group["comparable_pairs"],
+                "total_pairs": len(group["pairs"]), "item_count": len(group["item_ids"]),
+                "coverage": group["comparable_pairs"] / len(group["pairs"])}
+        if len(matches) > 1:
+            item["status"] = "duplicate_group"
+        elif len(matches) == 1:
+            entry = matches[0]
+            order = entry.get("order")
+            if (not isinstance(order, list) or any(not isinstance(x, str) for x in order)
+                  or len(order) != len(group["item_ids"])
+                  or len(order) != len(set(order)) or set(order) != set(group["item_ids"])):
+                item["status"] = "invalid_permutation"
+            else:
+                positions = {item_id: i for i, item_id in enumerate(order)}
+                pair_details = []
+                for pair in group["pairs"]:
+                    wrong = ((positions[pair["item_i"]] - positions[pair["item_j"]])
+                             * pair["mean_difference"]) < 0
+                    pair_details.append({**pair, "reversed": wrong,
+                                         "counted_inversion": int(wrong and pair["comparable"] == 1)})
+                lost = sum(pair["counted_inversion"] for pair in pair_details)
+                total = group["comparable_pairs"]
+                reward = 1.0 - lost / total
+                item.update(status="valid", group_pair_accuracy=max(0.0, min(1.0, reward)),
+                            order=order, inversion_pairs=lost, correct_pairs=total-lost,
+                            comparable_pairs=total, total_pairs=len(pair_details),
+                            coverage=total/len(pair_details),
+                            reported_value_counted_pairs=sum(p["comparable"] for p in pair_details
+                                                             if p["comparison_basis"] == "reported_value"),
+                            uncertainty_adjusted_counted_pairs=sum(p["comparable"] for p in pair_details
+                                                                  if p["comparison_basis"] == "uncertainty_adjusted"),
+                            all_reported_value_unequal_pairs=sum(p["mean_difference"] != 0 for p in pair_details),
+                            all_reported_value_correct_pairs=sum(p["mean_difference"] != 0 and not p["reversed"]
+                                                                 for p in pair_details),
+                            pairs=pair_details)
+        details.append(item)
+    return {"task_id": contract["task_id"], "task_version": contract["task_version"],
+            "task_pair_accuracy": aggregate_accuracy(details),
+            "random_q95": q,
+            "ranking_reward": reward_from_accuracy(aggregate_accuracy(details), q),
+            "submission_status": "valid" if all(g["status"] == "valid" for g in details) else "group_errors",
+            "groups": details, "official_reward": None,
+            "review_status": "not_reviewed", "evaluation_status": "completed"}
+
+
+def read_regular(path, limit):
+    try:
+        fd = os.open(path, os.O_RDONLY | os.O_NOFOLLOW | os.O_NONBLOCK)
+        try:
+            info = os.fstat(fd)
+            if not stat.S_ISREG(info.st_mode):
+                raise SubmissionError("not_regular_file")
+            if info.st_size > limit:
+                raise SubmissionError("submission_too_large")
+            with os.fdopen(fd, "rb", closefd=False) as stream:
+                raw = stream.read(limit + 1)
+            if len(raw) > limit:
+                raise SubmissionError("submission_too_large")
+            return raw
+        finally:
+            os.close(fd)
+    except OSError as error:
+        if error.errno in (errno.ENOENT, errno.ELOOP, errno.EISDIR, errno.ENOTDIR,
+                           errno.ENXIO, errno.EACCES, errno.EPERM):
+            raise SubmissionError("missing_or_unreadable_submission") from error
+        raise
+
+
+def grade_file(contract, answers, path, baseline):
+    validate_task(contract, answers)
+    q = validate_baseline(contract, answers, baseline)
+    try:
+        raw = read_regular(path, contract["max_submission_bytes"])
+        submission = load_json(raw)
+    except (SubmissionError, UnicodeError, json.JSONDecodeError, RecursionError) as error:
+        report = zero_report(contract, str(error) or type(error).__name__)
+        report["random_q95"] = q
+        return report
+    report = grade_submission(contract, answers, submission, baseline)
+    report["random_q95"] = q
+    report["submission_sha256"] = hashlib.sha256(raw).hexdigest()
+    return report
+
+
+def dump(path, value):
+    path.write_text(json.dumps(value, indent=2, allow_nan=False) + "\n", encoding="utf-8")
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--contract", type=Path, default=Path("/tests/contract.json"))
+    parser.add_argument("--answers", type=Path, default=Path("/tests/answers.json"))
+    parser.add_argument("--baseline", type=Path, default=Path("/tests/random_baseline.json"))
+    parser.add_argument("--submission", type=Path, default=Path("/logs/artifacts/rankings.json"))
+    parser.add_argument("--output-dir", type=Path, default=Path("/logs/verifier"))
+    args = parser.parse_args()
+    args.output_dir.mkdir(parents=True, exist_ok=True)
+    for name in ("reward.txt", "reward.json", "report.json", "evaluator_error.json"):
+        (args.output_dir / name).unlink(missing_ok=True)
+    try:
+        contract = load_json(args.contract.read_bytes())
+        answers = load_json(args.answers.read_bytes())
+        baseline = load_json(args.baseline.read_bytes())
+        report = grade_file(contract, answers, args.submission, baseline)
+    except (TaskError, SubmissionError, OSError, ValueError, KeyError, TypeError) as error:
+        dump(args.output_dir / "evaluator_error.json", {"status": "task_error", "error": str(error)})
+        raise SystemExit(2) from error
+    dump(args.output_dir / "report.json", report)
+    (args.output_dir / "reward.txt").write_text(str(report["ranking_reward"]) + "\n", encoding="utf-8")
+    print(json.dumps({"ranking_reward": report["ranking_reward"],
+                      "submission_status": report["submission_status"],
+                      "task_pair_accuracy": report["task_pair_accuracy"],
+                      "random_q95": report["random_q95"],
+                      "aggregation_zero_due_to_format": report.get("aggregation_zero_due_to_format", False),
+                      "official_reward": None, "review_status": "not_reviewed", "evaluation_status": "completed"}))
+
+
+if __name__ == "__main__":
+    main()

tasks/agcute-bending-endpoints-ranking/tests/random_baseline.json

@@ -0,0 +1,66 @@
+{
+  "method": "uniform_complete_permutations",
+  "generator": "numpy.random.PCG64",
+  "numpy_version": "2.4.4",
+  "quantile": 0.95,
+  "quantile_method": "higher",
+  "calibration_count": 200000,
+  "validation_count": 200000,
+  "calibration_seed": 17436290277762341355,
+  "validation_seed": 7360250039651054117,
+  "q": 0.6527777777777778,
+  "task_id": "agcute-bending-endpoints-ranking",
+  "task_version": "1.0.0",
+  "group_order": [
+    "max_stress",
+    "strain_at_max"
+  ],
+  "item_order": {
+    "max_stress": [
+      "exp_65960918c5e1",
+      "exp_0b361eedc092",
+      "exp_0a801637748b",
+      "exp_39fd1eed6b1c",
+      "exp_365507fb236c",
+      "exp_3c3c8522da9d",
+      "exp_fc4984656462",
+      "exp_3fed9f8f07c9",
+      "exp_aaa9e2c6f5b9"
+    ],
+    "strain_at_max": [
+      "exp_54b8fc8b3571",
+      "exp_3aa8fedf978b",
+      "exp_2d9d2ad06f6e",
+      "exp_8cfb0d14a259",
+      "exp_d53afffea8ad",
+      "exp_69a2a5342ec6",
+      "exp_eb05cc2a9dc8",
+      "exp_7b741589c10a",
+      "exp_4369fc61d66d"
+    ]
+  },
+  "calibration_code_sha256": "cbb437c19ae80d85f28374fe62c71e8052975ddc48c447fe89b9fb9913e8c6b1",
+  "relations_sha256": "c6be6dea1116a1cb8b9f113166040f1e0f907cb91946375bfcbe2578804943bc",
+  "weights_sha256": "e2f8f70a28b53aadf0d15ba87db844c29814b1f2511f1fbedb4f15da3007c20e",
+  "validation": {
+    "task_pair_accuracy": {
+      "mean": 0.4998954166666667,
+      "mean_standard_error": 0.00021030780229321986
+    },
+    "task_spearman": {
+      "mean": -1.5541666666666494e-05,
+      "mean_standard_error": 0.0005583706707422961
+    },
+    "ranking_reward": {
+      "mean": 0.005649400000000001,
+      "mean_standard_error": 7.235473429042408e-05
+    },
+    "zero_reward_fraction": 0.95489
+  },
+  "bindings": {
+    "contract_sha256": "f584d2b0ec67e9f047bc763fce2e0afc42a9fe4a2676db8c826dfec8dabc4ae2",
+    "answers_sha256": "beb49bf263eccb3f24b4bf2f0eddbc0d255a358b34418a20e858708ccf304e19",
+    "grader_sha256": "82c9d3b9c1fa5c2e2a39852e315df9e04714dee2ffa9eb64373014d8b4aeb381",
+    "auxiliary_grader_sha256": "b91d5af76c52dca0cb7195777a0cbea690a91af15f6bdbb0b13714e03ab45492"
+  }
+}

tasks/agcute-bending-endpoints-ranking/tests/test.sh

@@ -0,0 +1,3 @@
+#!/bin/sh
+set -eu
+exec python /tests/grade.py --root /tests --submission /logs/artifacts/rankings.json --output /logs/verifier

tasks/alkene-transfer-conditions-ranking/environment/Dockerfile

@@ -0,0 +1,7 @@
+FROM python@sha256:78387bc3881b8273120a12ebe6c1ab22b018ccc2c9adf565ae1ac9b536e184ea
+ENV PYTHONDONTWRITEBYTECODE=1 PYTHONUNBUFFERED=1
+RUN useradd -m -u 1000 -s /bin/bash solver && mkdir -p /logs/artifacts && chmod 777 /logs/artifacts
+WORKDIR /workspace
+COPY public/ /workspace/public/
+ENV OPENBLAS_NUM_THREADS=8 OMP_NUM_THREADS=8 MKL_NUM_THREADS=8
+USER solver

tasks/alkene-transfer-conditions-ranking/environment/public/contract.json

@@ -0,0 +1,111 @@
+{
+  "task_id": "alkene-transfer-conditions-ranking",
+  "task_version": "1.0.0",
+  "groups": [
+    {
+      "group_id": "g1",
+      "label": "4-Phenyl-1-butene cross-metathesis adduct yield",
+      "item_ids": [
+        "exp_41bd98dba742",
+        "exp_484d7efaebd5",
+        "exp_389c830f7e89",
+        "exp_dd127eb2b1f5",
+        "exp_8545262bd121",
+        "exp_369ffbca2e28",
+        "exp_cac98fcd26c1"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    },
+    {
+      "group_id": "g2",
+      "label": "Rotenone cross-metathesis adduct yield",
+      "item_ids": [
+        "exp_066e33c896b9",
+        "exp_fba92a14e6bb",
+        "exp_9279aa7cd389",
+        "exp_8e5983854cd1",
+        "exp_343b2ac1c9a7",
+        "exp_e01ea673fa5d",
+        "exp_ea300526f709"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    },
+    {
+      "group_id": "g3",
+      "label": "Combined alkene yield from E/Z latent precursors",
+      "item_ids": [
+        "exp_529d2754aa6a",
+        "exp_94c775c85236"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    }
+  ],
+  "scoring_method": "spearman",
+  "primary_score": "task_spearman",
+  "spearman_range": [
+    -1,
+    1
+  ],
+  "spearman_tie_method": "average",
+  "spearman_undefined_policy": "task_error",
+  "spearman_aggregation": "Pearson correlation of true average ranks and submitted positions, separately within each group, then fixed group-weighted sum; do not pool groups or clip negative groups.",
+  "secondary_scoring_method": "equal_pair_inversion_random95",
+  "ranking_reward_range": [
+    0,
+    1
+  ],
+  "random_quantile": 0.95,
+  "unknown_uncertainty_policy": "use_reported_value",
+  "pair_rule": "For each unordered pair, count it iff abs(mu_i-mu_j)>k*s_ij+epsilon for known uncertainty, or >epsilon for unknown uncertainty. Each counted pair has equal weight within its group. Unknown uncertainty is not zero uncertainty.",
+  "ranking_reward_formula": "Validate finite 0.5<=q<1 and 0<=A<=1; if A==1 return 1; else if A<=q+1e-12 return 0; otherwise clamp((A-q)/(1-q),0,1). A is the fixed weighted sum of group pair accuracies; q is the whole-task random 95th percentile.",
+  "format_errors": {
+    "whole_task": "Core=-1; auxiliary=0 for malformed/oversized/non-UTF8 JSON, duplicate object keys, wrong/missing identity, invalid top-level/rankings, unknown or unassignable group.",
+    "known_group": "Missing/duplicate known group or invalid item permutation: raw group Spearman=null, aggregation value=-1, group pair accuracy=0; preserve group weight and apply auxiliary threshold after aggregation."
+  },
+  "max_submission_bytes": 1048576,
+  "minimum_conditions_per_group": 2,
+  "time_limit_sec": 28800,
+  "solver_resources": {
+    "cpus": 8,
+    "memory_mb": 16384,
+    "gpus": 1,
+    "gpu_types": [
+      "H20"
+    ]
+  },
+  "submission_path": "/logs/artifacts/rankings.json",
+  "report_path": "/logs/artifacts/report.md",
+  "sources_path": "/logs/artifacts/sources.json",
+  "analysis_path": "/logs/artifacts/analysis/",
+  "forbidden_result_scope": "Measured outcomes, yields, rankings, equivalent endpoints and result-fitted predictors for alkene transfer-reagent and catalyst conditions in the stated material and reaction system are prohibited."
+}

tasks/alkene-transfer-conditions-ranking/environment/public/experiment.md

@@ -0,0 +1,58 @@
+# Experimental definition
+
+Compare the listed experimental conditions for alkene transfer-reagent and catalyst conditions. The inputs define the materials, treatment and measured endpoint.
+
+## Common conditions
+
+| Field | Setting |
+| --- | --- |
+| material_handling | `Oven-dried glassware underN2 usingSchlenk techniques.All specified solvents are anhydrous.Commercial reagents used as supplied;one-carbon-transfer reagents stored in freezer underN2,numeric storage temperature unspecified.` |
+| reagents | `Transfer reagent1a:((allylsulfonyl)methoxy)(tert-butyl)dimethylsilane,SMILES C=CCS(=O)(=O)CO[Si](C)(C)C(C)(C)C. Transfer reagent1b:tert-butyldimethyl(((3-methylbut-2-en-1-yl)sulfonyl)methoxy)silane,SMILES CC(C)=CCS(=O)(=O)CO[Si](C)(C)C(C)(C)C. TBS means tert-butyldimethylsilyl.` |
+| endpoint_convention | `Rank measured product1H NMR yield within each comparison group. Cross-metathesis groups concern total latent adduct; the E/Z-reactant group concerns the combined alkene isomers after fragmentation. Do not combine these endpoints across groups.` |
+
+## Procedure and measurement
+
+Combine each listed condition with the common procedure. Rank the specified measured endpoint separately within each comparison group.
+
+- `g1` — product 1H NMR yield (%): Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard.
+- `g2` — product 1H NMR yield (%): Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard.
+- `g3` — product 1H NMR yield (%): Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard.
+
+## Item settings
+
+Expand each item in targets.csv or protocol.json together with the common conditions. Every selected item is listed using the same fields.
+
+## Source limitations relevant to the definition
+
+- Independent replicate counts and numerical uncertainty are not specified.
+- The source reports NMR point yields without numerical uncertainty. E/Z product fractions are not used as the combined-yield endpoint.
+
+### structures/metathesis_catalysts.png
+
+HG-II,G-II andG-I catalyst molecular structures.
+
+![Experimental structure](structures/metathesis_catalysts.png)
+
+### structures/rotenone_reaction.png
+
+Rotenone stereostructure,transfer reagents1a/1b,andcross-metathesis product.
+
+![Experimental structure](structures/rotenone_reaction.png)
+
+### structures/rotenone_catalysts.png
+
+Catalyst molecular structures includingnitro-GrelaI2SiPr.
+
+![Experimental structure](structures/rotenone_catalysts.png)
+
+### structures/latent_E.png
+
+E-configured latent precursorE-17;Bn=benzyl,TBS=tert-butyldimethylsilyl.
+
+![Experimental structure](structures/latent_E.png)
+
+### structures/latent_Z.png
+
+Z-configured latent precursorZ-17;Bn=benzyl,TBS=tert-butyldimethylsilyl.
+
+![Experimental structure](structures/latent_Z.png)

tasks/alkene-transfer-conditions-ranking/environment/public/information_policy.md

@@ -0,0 +1,12 @@
+# Information policy
+
+The original study, all its supplementary materials (including methods), source data, related reporting, summaries, mirrors and caches are prohibited. The same restriction covers external values, rankings, qualitative comparisons, equivalent endpoints and predictors fitted to target or same-type experimental results. Measured outcomes, yields, rankings, equivalent endpoints and result-fitted predictors for alkene transfer-reagent and catalyst conditions in the stated material and reaction system are prohibited.
+
+Allowed examples:
+- General reaction mechanisms and physical chemistry independent of the prohibited outcome measurements.
+
+Forbidden examples:
+- Original study, supporting information, data and reproductions of these experiments.
+- Measured or inferred rankings calibrated on the same-type experimental outcomes.
+
+Use audited tools for all searches, downloads and code-initiated network requests. Check both requests and returned content before use. Do not circumvent access controls or ask another model/service for forbidden results. If a permitted query unexpectedly exposes prohibited external results, immediately stop prediction and further search; record only the triggering step, source and stopping reason in report.md without restating results. Do not retrieve or reconstruct specific experimental results from your own model memory; briefly disclose a noticed recollection in your report and continue using permitted inputs and reasoning.

tasks/alkene-transfer-conditions-ranking/environment/public/protocol.json

@@ -0,0 +1,267 @@
+{
+  "common_conditions": {
+    "material_handling": "Oven-dried glassware underN2 usingSchlenk techniques.All specified solvents are anhydrous.Commercial reagents used as supplied;one-carbon-transfer reagents stored in freezer underN2,numeric storage temperature unspecified.",
+    "reagents": "Transfer reagent1a:((allylsulfonyl)methoxy)(tert-butyl)dimethylsilane,SMILES C=CCS(=O)(=O)CO[Si](C)(C)C(C)(C)C. Transfer reagent1b:tert-butyldimethyl(((3-methylbut-2-en-1-yl)sulfonyl)methoxy)silane,SMILES CC(C)=CCS(=O)(=O)CO[Si](C)(C)C(C)(C)C. TBS means tert-butyldimethylsilyl.",
+    "endpoint_convention": "Rank measured product1H NMR yield within each comparison group. Cross-metathesis groups concern total latent adduct; the E/Z-reactant group concerns the combined alkene isomers after fragmentation. Do not combine these endpoints across groups."
+  },
+  "measurement": {
+    "name": "product 1H NMR yield",
+    "unit": "%",
+    "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+  },
+  "measurements_by_group": {
+    "g1": {
+      "name": "product 1H NMR yield",
+      "unit": "%",
+      "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+    },
+    "g2": {
+      "name": "product 1H NMR yield",
+      "unit": "%",
+      "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+    },
+    "g3": {
+      "name": "product 1H NMR yield",
+      "unit": "%",
+      "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+    }
+  },
+  "measurement_rule": "Use measurements_by_group for each group. The measurement field is a default summary and does not override a group-specific endpoint.",
+  "inheritance_rule": "common_conditions apply unless the same field is explicitly overridden by the item; item settings are literal experimental facts, not predicted outcomes.",
+  "items": {
+    "exp_41bd98dba742": {
+      "alkene": "4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 4,
+      "catalyst": "G-II as structures/metathesis_catalysts.png",
+      "catalyst_mol_percent": 5,
+      "solvent": "dichloromethane",
+      "solvent_volume_mL": 1,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 8,
+      "procedure": "5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.",
+      "analysis": "Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.",
+      "product": "PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_484d7efaebd5": {
+      "alkene": "4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 2,
+      "catalyst": "HG-II as structures/metathesis_catalysts.png",
+      "catalyst_mol_percent": 5,
+      "solvent": "dichloromethane",
+      "solvent_volume_mL": 1,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 8,
+      "procedure": "5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.",
+      "analysis": "Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.",
+      "product": "PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_389c830f7e89": {
+      "alkene": "4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 1,
+      "catalyst": "HG-II as structures/metathesis_catalysts.png",
+      "catalyst_mol_percent": 5,
+      "solvent": "dichloromethane",
+      "solvent_volume_mL": 1,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 8,
+      "procedure": "5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.",
+      "analysis": "Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.",
+      "product": "PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_dd127eb2b1f5": {
+      "alkene": "4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 4,
+      "catalyst": "G-I as structures/metathesis_catalysts.png",
+      "catalyst_mol_percent": 5,
+      "solvent": "dichloromethane",
+      "solvent_volume_mL": 1,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 8,
+      "procedure": "5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.",
+      "analysis": "Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.",
+      "product": "PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_8545262bd121": {
+      "alkene": "4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",
+      "transfer_reagent": "1b",
+      "transfer_reagent_equiv": 3,
+      "catalyst": "HG-II as structures/metathesis_catalysts.png",
+      "catalyst_mol_percent": 5,
+      "solvent": "toluene",
+      "solvent_volume_mL": 1,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 100,
+      "time_h": 8,
+      "procedure": "5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.",
+      "analysis": "Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.",
+      "product": "PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_369ffbca2e28": {
+      "alkene": "4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",
+      "transfer_reagent": "1b",
+      "transfer_reagent_equiv": 3,
+      "catalyst": "HG-II as structures/metathesis_catalysts.png",
+      "catalyst_mol_percent": 5,
+      "solvent": "dichloromethane",
+      "solvent_volume_mL": 1,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 8,
+      "procedure": "5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.",
+      "analysis": "Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.",
+      "product": "PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_cac98fcd26c1": {
+      "alkene": "4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 3,
+      "catalyst": "HG-II as structures/metathesis_catalysts.png",
+      "catalyst_mol_percent": 5,
+      "solvent": "dichloromethane",
+      "solvent_volume_mL": 1,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 8,
+      "procedure": "5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.",
+      "analysis": "Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.",
+      "product": "PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_066e33c896b9": {
+      "alkene": "Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",
+      "transfer_reagent": "1b",
+      "transfer_reagent_equiv": 3,
+      "catalyst": "HG-II as structures/rotenone_catalysts.png",
+      "catalyst_mol_percent": 20,
+      "solvent": "1,2-dichloroethane",
+      "solvent_volume_mL": 0.5,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 100,
+      "time_h": 16,
+      "procedure": "10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.",
+      "analysis": "Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.",
+      "product": "Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_fba92a14e6bb": {
+      "alkene": "Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 10,
+      "catalyst": "HG-II as structures/rotenone_catalysts.png",
+      "catalyst_mol_percent": 20,
+      "solvent": "1,2-dichloroethane",
+      "solvent_volume_mL": 0.5,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 100,
+      "time_h": 5,
+      "procedure": "10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.",
+      "analysis": "Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.",
+      "product": "Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_9279aa7cd389": {
+      "alkene": "Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",
+      "transfer_reagent": "1b",
+      "transfer_reagent_equiv": 2,
+      "catalyst": "HG-II as structures/rotenone_catalysts.png",
+      "catalyst_mol_percent": 20,
+      "solvent": "1,2-dichloroethane",
+      "solvent_volume_mL": 0.5,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 100,
+      "time_h": 16,
+      "procedure": "10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.",
+      "analysis": "Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.",
+      "product": "Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_8e5983854cd1": {
+      "alkene": "Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 10,
+      "catalyst": "nitro-GrelaI2SiPr as structures/rotenone_catalysts.png",
+      "catalyst_mol_percent": 20,
+      "solvent": "1,2-dichloroethane",
+      "solvent_volume_mL": 0.5,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 100,
+      "time_h": 5,
+      "procedure": "10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.",
+      "analysis": "Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.",
+      "product": "Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_343b2ac1c9a7": {
+      "alkene": "Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 10,
+      "catalyst": "HG-II as structures/rotenone_catalysts.png",
+      "catalyst_mol_percent": 20,
+      "solvent": "1,2-dichloroethane",
+      "solvent_volume_mL": 0.5,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 100,
+      "time_h": 16,
+      "procedure": "10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.",
+      "analysis": "Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.",
+      "product": "Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_e01ea673fa5d": {
+      "alkene": "Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",
+      "transfer_reagent": "1b",
+      "transfer_reagent_equiv": 5,
+      "catalyst": "HG-II as structures/rotenone_catalysts.png",
+      "catalyst_mol_percent": 20,
+      "solvent": "1,2-dichloroethane",
+      "solvent_volume_mL": 0.5,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 100,
+      "time_h": 16,
+      "procedure": "10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.",
+      "analysis": "Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.",
+      "product": "Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_ea300526f709": {
+      "alkene": "Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",
+      "transfer_reagent": "1a",
+      "transfer_reagent_equiv": 10,
+      "catalyst": "HG-II as structures/rotenone_catalysts.png",
+      "catalyst_mol_percent": 20,
+      "solvent": "dichloromethane",
+      "solvent_volume_mL": 0.5,
+      "alkene_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 16,
+      "procedure": "10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.",
+      "analysis": "Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.",
+      "product": "Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation."
+    },
+    "exp_529d2754aa6a": {
+      "latent_precursor": "E-17,fullmolecularstructure in structures/latent_E.png,36.8mg0.100mmol",
+      "citric_acid": "57.6mg0.300mmol3equiv",
+      "fluoride": "Tetrabutylammoniumfluoride trihydrate39.4mg0.125mmol1.25equiv",
+      "solvent": "0.5mLdichloromethane plus0.5mLmethanol,total1.0mL",
+      "substrate_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 16,
+      "procedure": "10mLmicrowavevial withmagneticbar.Evacuate/refillN2threecycles.Adddichloromethane,heatin oilbath.Thenaddcitricacid andTBAFtrihydrate dissolved inmethanol throughseptum;heatandstir16h.",
+      "analysis": "Addmethylbenzoate12.6microlitre0.100mmol.Takecrudealiquot for1H NMR combinedproductyield;GC-MS separatelydeterminesisomerratio,butratiois notthe requestedendpoint.",
+      "product": "CombinedE+Z1-phenylhex-4-ene(C6H5CH2CH2CH2CH=CHCH3) afterfragmentation."
+    },
+    "exp_94c775c85236": {
+      "latent_precursor": "Z-17,fullmolecularstructure in structures/latent_Z.png,36.8mg0.100mmol",
+      "citric_acid": "57.6mg0.300mmol3equiv",
+      "fluoride": "Tetrabutylammoniumfluoride trihydrate39.4mg0.125mmol1.25equiv",
+      "solvent": "0.5mLdichloromethane plus0.5mLmethanol,total1.0mL",
+      "substrate_concentration_M": 0.1,
+      "temperature_C": 40,
+      "time_h": 16,
+      "procedure": "10mLmicrowavevial withmagneticbar.Evacuate/refillN2threecycles.Adddichloromethane,heatin oilbath.Thenaddcitricacid andTBAFtrihydrate dissolved inmethanol throughseptum;heatandstir16h.",
+      "analysis": "Addmethylbenzoate12.6microlitre0.100mmol.Takecrudealiquot for1H NMR combinedproductyield;GC-MS separatelydeterminesisomerratio,butratiois notthe requestedendpoint.",
+      "product": "CombinedE+Z1-phenylhex-4-ene(C6H5CH2CH2CH2CH=CHCH3) afterfragmentation."
+    }
+  }
+}

tasks/alkene-transfer-conditions-ranking/environment/public/resources.md

@@ -0,0 +1,3 @@
+# Resources
+
+You have 8 CPU cores, 16,384 MiB system RAM and one NVIDIA H20 GPU, shared by all local processes. Remote model APIs, databases and services do not consume the local compute quota; their calls and waits consume the same 28,800-second wall-clock budget. Record service identity, requests and usage. All compute and network activity must follow the information policy. The 8-hour clock starts at first delivery of the task inputs; retries do not reset it.

tasks/alkene-transfer-conditions-ranking/environment/public/targets.csv

@@ -0,0 +1,17 @@
+group_id,item_id,settings_ref,alkene,transfer_reagent,transfer_reagent_equiv,catalyst,catalyst_mol_percent,solvent,solvent_volume_mL,alkene_concentration_M,temperature_C,time_h,procedure,analysis,product,latent_precursor,citric_acid,fluoride,substrate_concentration_M
+g1,exp_41bd98dba742,protocol.json#exp_41bd98dba742,"4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",1a,4,G-II as structures/metathesis_catalysts.png,5,dichloromethane,1,0.1,40,8,"5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.","Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.","PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation.",,,,
+g1,exp_484d7efaebd5,protocol.json#exp_484d7efaebd5,"4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",1a,2,HG-II as structures/metathesis_catalysts.png,5,dichloromethane,1,0.1,40,8,"5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.","Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.","PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation.",,,,
+g1,exp_389c830f7e89,protocol.json#exp_389c830f7e89,"4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",1a,1,HG-II as structures/metathesis_catalysts.png,5,dichloromethane,1,0.1,40,8,"5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.","Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.","PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation.",,,,
+g1,exp_dd127eb2b1f5,protocol.json#exp_dd127eb2b1f5,"4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",1a,4,G-I as structures/metathesis_catalysts.png,5,dichloromethane,1,0.1,40,8,"5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.","Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.","PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation.",,,,
+g1,exp_8545262bd121,protocol.json#exp_8545262bd121,"4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",1b,3,HG-II as structures/metathesis_catalysts.png,5,toluene,1,0.1,100,8,"5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.","Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.","PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation.",,,,
+g1,exp_369ffbca2e28,protocol.json#exp_369ffbca2e28,"4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",1b,3,HG-II as structures/metathesis_catalysts.png,5,dichloromethane,1,0.1,40,8,"5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.","Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.","PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation.",,,,
+g1,exp_cac98fcd26c1,protocol.json#exp_cac98fcd26c1,"4-phenyl-1-butene,C=CCCc1ccccc1,15microlitre0.100mmol1equiv",1a,3,HG-II as structures/metathesis_catalysts.png,5,dichloromethane,1,0.1,40,8,"5mLround-bottomflask,magneticbar,refluxcondenser withTeflontapedjoint,suba-seal,N2line andoutletneedle toventethylene.Heat reagents in oilbath,add0.25mLsolvent,0.5mLcatalyststock then0.25mLsolvent.","Evaporate solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate12.6microlitre0.100mmol.","PhCH2CH2CH=CHCH2S(=O)2CH2OTBS,totalgeometricisomers,beforefragmentation.",,,,
+g2,exp_066e33c896b9,protocol.json#exp_066e33c896b9,"Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",1b,3,HG-II as structures/rotenone_catalysts.png,20,"1,2-dichloroethane",0.5,0.1,100,16,"10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.","Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.","Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation.",,,,
+g2,exp_fba92a14e6bb,protocol.json#exp_fba92a14e6bb,"Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",1a,10,HG-II as structures/rotenone_catalysts.png,20,"1,2-dichloroethane",0.5,0.1,100,5,"10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.","Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.","Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation.",,,,
+g2,exp_9279aa7cd389,protocol.json#exp_9279aa7cd389,"Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",1b,2,HG-II as structures/rotenone_catalysts.png,20,"1,2-dichloroethane",0.5,0.1,100,16,"10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.","Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.","Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation.",,,,
+g2,exp_8e5983854cd1,protocol.json#exp_8e5983854cd1,"Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",1a,10,nitro-GrelaI2SiPr as structures/rotenone_catalysts.png,20,"1,2-dichloroethane",0.5,0.1,100,5,"10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.","Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.","Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation.",,,,
+g2,exp_343b2ac1c9a7,protocol.json#exp_343b2ac1c9a7,"Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",1a,10,HG-II as structures/rotenone_catalysts.png,20,"1,2-dichloroethane",0.5,0.1,100,16,"10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.","Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.","Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation.",,,,
+g2,exp_e01ea673fa5d,protocol.json#exp_e01ea673fa5d,"Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",1b,5,HG-II as structures/rotenone_catalysts.png,20,"1,2-dichloroethane",0.5,0.1,100,16,"10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.","Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.","Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation.",,,,
+g2,exp_ea300526f709,protocol.json#exp_ea300526f709,"Rotenone,stereostructure shown in structures/rotenone_reaction.png,19.7mg0.050mmol1equiv",1a,10,HG-II as structures/rotenone_catalysts.png,20,dichloromethane,0.5,0.1,40,16,"10mLmicrowavevial withmagneticbar.Charge reagents andcatalyst,crimpseal,evacuate/refillN2threecycles,addsolventviasyringe.Heatstir,thenquenchcatalyst with~0.1mLbutylvinylether.","Remove solvent in vacuo,measure1H NMR adduct yield againstmethylbenzoate6.3microlitre0.050mmol.","Rotenone cross-metathesis product in structures/rotenone_reaction.png,totalgeometricisomers,beforefragmentation.",,,,
+g3,exp_529d2754aa6a,protocol.json#exp_529d2754aa6a,,,,,,"0.5mLdichloromethane plus0.5mLmethanol,total1.0mL",,,40,16,"10mLmicrowavevial withmagneticbar.Evacuate/refillN2threecycles.Adddichloromethane,heatin oilbath.Thenaddcitricacid andTBAFtrihydrate dissolved inmethanol throughseptum;heatandstir16h.","Addmethylbenzoate12.6microlitre0.100mmol.Takecrudealiquot for1H NMR combinedproductyield;GC-MS separatelydeterminesisomerratio,butratiois notthe requestedendpoint.",CombinedE+Z1-phenylhex-4-ene(C6H5CH2CH2CH2CH=CHCH3) afterfragmentation.,"E-17,fullmolecularstructure in structures/latent_E.png,36.8mg0.100mmol",57.6mg0.300mmol3equiv,Tetrabutylammoniumfluoride trihydrate39.4mg0.125mmol1.25equiv,0.1
+g3,exp_94c775c85236,protocol.json#exp_94c775c85236,,,,,,"0.5mLdichloromethane plus0.5mLmethanol,total1.0mL",,,40,16,"10mLmicrowavevial withmagneticbar.Evacuate/refillN2threecycles.Adddichloromethane,heatin oilbath.Thenaddcitricacid andTBAFtrihydrate dissolved inmethanol throughseptum;heatandstir16h.","Addmethylbenzoate12.6microlitre0.100mmol.Takecrudealiquot for1H NMR combinedproductyield;GC-MS separatelydeterminesisomerratio,butratiois notthe requestedendpoint.",CombinedE+Z1-phenylhex-4-ene(C6H5CH2CH2CH2CH=CHCH3) afterfragmentation.,"Z-17,fullmolecularstructure in structures/latent_Z.png,36.8mg0.100mmol",57.6mg0.300mmol3equiv,Tetrabutylammoniumfluoride trihydrate39.4mg0.125mmol1.25equiv,0.1

tasks/alkene-transfer-conditions-ranking/instruction.md

@@ -0,0 +1,222 @@
+You are a general-purpose AI scientist tasked with predicting experimental outcomes across scientific disciplines. Your goal is to produce the most accurate, scientifically grounded predictions possible from the information provided and any permitted information you can obtain. For this task, predict only the ordering of experimental outcomes within each specified comparison group. Absolute outcome values are not required.
+
+You may use all computational tools and search resources available in your environment, including web search, scientific literature, databases, code execution, statistical modeling, machine learning, and numerical simulation. Select and combine methods appropriate to the task, and actively perform calculations when they can improve your predictions. Proactively check the available CPUs, GPUs, and software, and identify key prediction problems they could help address; whether to use a GPU should depend on the suitability of the method. For scientifically grounded methods that may be useful but whose benefits are uncertain, prioritize small-scale trial calculations before deciding whether to expand the computation. You may download any software, code, models, and data you consider necessary from the internet, subject to the following restrictions: do not search for, access, or use the original study, its supplementary materials, or data that can directly reveal the experimental results, and do not use the results of the target experiments or experiments of the same type.
+
+Ground your analysis in the experimental setup, measurement definitions, controls, and relevant scientific mechanisms. Gather supporting evidence from permitted sources, assess its applicability to the specific experiments, and evaluate alternative explanations. Check assumptions, units, and consistency, and refine your predictions as new permitted evidence becomes available. Use general scientific knowledge to make new predictions. Do not retrieve, reconstruct, or use specific experimental results or rankings from your model memory or training data. After forming an initial ranking, continue to look for and actually try methods that could improve key judgments within the time budget, prioritizing resources for the most uncertain comparisons.
+
+Clearly distinguish experimental observations, computational results, and scientific inference. Assess uncertainty and methodological limitations. Do not fabricate sources, data, tool outputs, or calculations, and do not present computational proxies as experimentally measured values.
+
+Follow the task-specific instructions and output format exactly. When explanations are permitted, provide a concise summary of your methods, key computational results, and supporting rationale. Keep supporting analysis separate from any final submission that requires a restricted format.
+
+You have a strict time limit of 8 hours (28,800 seconds). Submit the required files before the deadline.
+
+# Alkene transfer-reagent and catalyst conditions
+
+## Prediction question
+
+Compare the listed experimental conditions for alkene transfer-reagent and catalyst conditions. The inputs define the materials, treatment and measured endpoint.
+
+Rank each comparison group from lowest to highest using that group's specified endpoint and unit. Submit rankings only; absolute values are not required.
+
+## Experimental definition
+
+Read `/workspace/public/experiment.md`, `protocol.json` and `targets.csv`. The common conditions in protocol.json apply to every item unless that item's setting explicitly overrides the same field. Expand both parts together. Public sequences, structures and other supplied assets are part of the definition. Do not obtain missing details from the prohibited original study.
+
+Combine each listed condition with the common procedure. Rank the specified measured endpoint separately within each comparison group.
+
+Outcome definitions by group:
+
+- `g1` — product 1H NMR yield (%): Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard.
+- `g2` — product 1H NMR yield (%): Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard.
+- `g3` — product 1H NMR yield (%): Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard.
+
+## Comparison groups
+
+- `g1`: 4-Phenyl-1-butene cross-metathesis adduct yield; 7 items; weight 0.3333333333333333; ascending product 1H NMR yield (%).
+- `g2`: Rotenone cross-metathesis adduct yield; 7 items; weight 0.3333333333333333; ascending product 1H NMR yield (%).
+- `g3`: Combined alkene yield from E/Z latent precursors; 2 items; weight 0.3333333333333333; ascending product 1H NMR yield (%).
+
+## Allowed and prohibited information
+
+The original study, all its supplementary materials (including methods), source data, related reporting, summaries, mirrors and caches are prohibited. The same restriction covers external values, rankings, qualitative comparisons, equivalent endpoints and predictors fitted to target or same-type experimental results. Measured outcomes, yields, rankings, equivalent endpoints and result-fitted predictors for alkene transfer-reagent and catalyst conditions in the stated material and reaction system are prohibited.
+
+Allowed examples:
+- General reaction mechanisms and physical chemistry independent of the prohibited outcome measurements.
+
+Forbidden examples:
+- Original study, supporting information, data and reproductions of these experiments.
+- Measured or inferred rankings calibrated on the same-type experimental outcomes.
+
+Use audited tools for all searches, downloads and code-initiated network requests. Check both requests and returned content before use. Do not circumvent access controls or ask another model/service for forbidden results. If a permitted query unexpectedly exposes prohibited external results, immediately stop prediction and further search; record only the triggering step, source and stopping reason in report.md without restating results. Do not retrieve or reconstruct specific experimental results from your own model memory; briefly disclose a noticed recollection in your report and continue using permitted inputs and reasoning.
+
+## Tools, resources and deadline
+
+You have 8 CPU cores, 16,384 MiB system RAM and one NVIDIA H20 GPU, shared by all local processes. Remote model APIs, databases and services do not consume the local compute quota; their calls and waits consume the same 28,800-second wall-clock budget. Record service identity, requests and usage. All compute and network activity must follow the information policy. The 8-hour clock starts at first delivery of the task inputs; retries do not reset it.
+
+## Submission
+
+Write UTF-8 JSON to `/logs/artifacts/rankings.json` (at most 1 MiB, unique object keys). Include the exact task identity and exactly one entry per group; each order must contain every assigned item once, from lowest to highest. The following is a format example in public display order, not a predicted answer:
+
+```json
+{
+  "task_id": "alkene-transfer-conditions-ranking",
+  "task_version": "1.0.0",
+  "rankings": [
+    {
+      "group_id": "g1",
+      "order": [
+        "exp_41bd98dba742",
+        "exp_484d7efaebd5",
+        "exp_389c830f7e89",
+        "exp_dd127eb2b1f5",
+        "exp_8545262bd121",
+        "exp_369ffbca2e28",
+        "exp_cac98fcd26c1"
+      ]
+    },
+    {
+      "group_id": "g2",
+      "order": [
+        "exp_066e33c896b9",
+        "exp_fba92a14e6bb",
+        "exp_9279aa7cd389",
+        "exp_8e5983854cd1",
+        "exp_343b2ac1c9a7",
+        "exp_e01ea673fa5d",
+        "exp_ea300526f709"
+      ]
+    },
+    {
+      "group_id": "g3",
+      "order": [
+        "exp_529d2754aa6a",
+        "exp_94c775c85236"
+      ]
+    }
+  ]
+}
+```
+
+Write an English account of scientific reasoning, calculations, assumptions and uncertainty to `/logs/artifacts/report.md`; source records to `/logs/artifacts/sources.json`; supporting scripts and calculations to `/logs/artifacts/analysis/`. The evaluator uses the last submission received before the deadline. Empty or missing submissions receive core -1 and auxiliary 0.
+
+## Ranking scores
+
+The primary score is standard Spearman rank correlation, from -1 to 1, computed separately in each group and combined with fixed group weights. Equal measured values receive average true ranks; predictions remain strict permutations. Correct/reverse rankings score 1/-1 when true values have no ties; with true ties the best strict ranking can score less than 1. Random complete rankings have expected correlation 0.
+
+The auxiliary score counts each decidable pair equally within its group, combines group accuracies using the same weights, then applies one whole-task random 95th-percentile threshold. Known uncertainty uses the declared difference-error rule; unknown uncertainty uses reported values with epsilon and does not exclude unequal results. The threshold is hidden. Core and auxiliary scores are recorded separately; neither meaning of zero proves the absence of predictive ability. Detailed algorithms and all format penalties are in the following authoritative public contract, from which the companion files are generated.
+
+<!-- public-contract -->
+```json
+{
+  "task_id": "alkene-transfer-conditions-ranking",
+  "task_version": "1.0.0",
+  "groups": [
+    {
+      "group_id": "g1",
+      "label": "4-Phenyl-1-butene cross-metathesis adduct yield",
+      "item_ids": [
+        "exp_41bd98dba742",
+        "exp_484d7efaebd5",
+        "exp_389c830f7e89",
+        "exp_dd127eb2b1f5",
+        "exp_8545262bd121",
+        "exp_369ffbca2e28",
+        "exp_cac98fcd26c1"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    },
+    {
+      "group_id": "g2",
+      "label": "Rotenone cross-metathesis adduct yield",
+      "item_ids": [
+        "exp_066e33c896b9",
+        "exp_fba92a14e6bb",
+        "exp_9279aa7cd389",
+        "exp_8e5983854cd1",
+        "exp_343b2ac1c9a7",
+        "exp_e01ea673fa5d",
+        "exp_ea300526f709"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    },
+    {
+      "group_id": "g3",
+      "label": "Combined alkene yield from E/Z latent precursors",
+      "item_ids": [
+        "exp_529d2754aa6a",
+        "exp_94c775c85236"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    }
+  ],
+  "scoring_method": "spearman",
+  "primary_score": "task_spearman",
+  "spearman_range": [
+    -1,
+    1
+  ],
+  "spearman_tie_method": "average",
+  "spearman_undefined_policy": "task_error",
+  "spearman_aggregation": "Pearson correlation of true average ranks and submitted positions, separately within each group, then fixed group-weighted sum; do not pool groups or clip negative groups.",
+  "secondary_scoring_method": "equal_pair_inversion_random95",
+  "ranking_reward_range": [
+    0,
+    1
+  ],
+  "random_quantile": 0.95,
+  "unknown_uncertainty_policy": "use_reported_value",
+  "pair_rule": "For each unordered pair, count it iff abs(mu_i-mu_j)>k*s_ij+epsilon for known uncertainty, or >epsilon for unknown uncertainty. Each counted pair has equal weight within its group. Unknown uncertainty is not zero uncertainty.",
+  "ranking_reward_formula": "Validate finite 0.5<=q<1 and 0<=A<=1; if A==1 return 1; else if A<=q+1e-12 return 0; otherwise clamp((A-q)/(1-q),0,1). A is the fixed weighted sum of group pair accuracies; q is the whole-task random 95th percentile.",
+  "format_errors": {
+    "whole_task": "Core=-1; auxiliary=0 for malformed/oversized/non-UTF8 JSON, duplicate object keys, wrong/missing identity, invalid top-level/rankings, unknown or unassignable group.",
+    "known_group": "Missing/duplicate known group or invalid item permutation: raw group Spearman=null, aggregation value=-1, group pair accuracy=0; preserve group weight and apply auxiliary threshold after aggregation."
+  },
+  "max_submission_bytes": 1048576,
+  "minimum_conditions_per_group": 2,
+  "time_limit_sec": 28800,
+  "solver_resources": {
+    "cpus": 8,
+    "memory_mb": 16384,
+    "gpus": 1,
+    "gpu_types": [
+      "H20"
+    ]
+  },
+  "submission_path": "/logs/artifacts/rankings.json",
+  "report_path": "/logs/artifacts/report.md",
+  "sources_path": "/logs/artifacts/sources.json",
+  "analysis_path": "/logs/artifacts/analysis/",
+  "forbidden_result_scope": "Measured outcomes, yields, rankings, equivalent endpoints and result-fitted predictors for alkene transfer-reagent and catalyst conditions in the stated material and reaction system are prohibited."
+}
+```
+<!-- /public-contract -->

tasks/alkene-transfer-conditions-ranking/solution/solve.sh

@@ -0,0 +1,42 @@
+#!/bin/sh
+set -eu
+mkdir -p /logs/artifacts
+cat > /logs/artifacts/rankings.json <<'PROJECTAA_ORACLE'
+{
+  "task_id": "alkene-transfer-conditions-ranking",
+  "task_version": "1.0.0",
+  "rankings": [
+    {
+      "group_id": "g1",
+      "order": [
+        "exp_dd127eb2b1f5",
+        "exp_369ffbca2e28",
+        "exp_8545262bd121",
+        "exp_389c830f7e89",
+        "exp_484d7efaebd5",
+        "exp_cac98fcd26c1",
+        "exp_41bd98dba742"
+      ]
+    },
+    {
+      "group_id": "g2",
+      "order": [
+        "exp_8e5983854cd1",
+        "exp_343b2ac1c9a7",
+        "exp_ea300526f709",
+        "exp_fba92a14e6bb",
+        "exp_9279aa7cd389",
+        "exp_e01ea673fa5d",
+        "exp_066e33c896b9"
+      ]
+    },
+    {
+      "group_id": "g3",
+      "order": [
+        "exp_94c775c85236",
+        "exp_529d2754aa6a"
+      ]
+    }
+  ]
+}
+PROJECTAA_ORACLE

tasks/alkene-transfer-conditions-ranking/task.toml

@@ -0,0 +1,27 @@
+schema_version = "1.2"
+[metadata]
+task_id = "alkene-transfer-conditions-ranking"
+task_version = "1.0.0"
+pipeline_version = "19"
+scientist_prompt_version = "projectaa-scientist-en-v7-ranking"
+[agent]
+user = "solver"
+timeout_sec = 28800.0
+[environment]
+build_timeout_sec = 900.0
+cpus = 8
+memory_mb = 16384
+storage_mb = 20480
+gpus = 1
+gpu_types = ["H20"]
+allow_internet = true
+[verifier]
+timeout_sec = 120.0
+environment_mode = "separate"
+[verifier.environment]
+build_timeout_sec = 600.0
+cpus = 1
+memory_mb = 1024
+storage_mb = 2048
+gpus = 0
+allow_internet = false

tasks/alkene-transfer-conditions-ranking/tests/Dockerfile

@@ -0,0 +1,5 @@
+FROM python@sha256:78387bc3881b8273120a12ebe6c1ab22b018ccc2c9adf565ae1ac9b536e184ea
+ENV PYTHONDONTWRITEBYTECODE=1 PYTHONUNBUFFERED=1
+COPY . /tests/
+RUN chmod +x /tests/test.sh
+WORKDIR /tests

tasks/alkene-transfer-conditions-ranking/tests/answers.json

@@ -0,0 +1,644 @@
+{
+  "task_id": "alkene-transfer-conditions-ranking",
+  "task_version": "1.0.0",
+  "groups": [
+    {
+      "group_id": "g1",
+      "items": [
+        {
+          "item_id": "exp_41bd98dba742",
+          "mean": 97,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p6 TableS3 entry2,catalyst structures;p5section3.1method",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_484d7efaebd5",
+          "mean": 89,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p5 TableS1 entry2,scheme andsection3.1method",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_389c830f7e89",
+          "mean": 66,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p5 TableS1 entry1,scheme andsection3.1method",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_dd127eb2b1f5",
+          "mean": 11,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p6 TableS3 entry3,catalyst structures;p5section3.1method",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_8545262bd121",
+          "mean": 53,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p6 TableS4 entry3,scheme;p5section3.1method",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_369ffbca2e28",
+          "mean": 53,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p6 TableS4 entry2,scheme;p5section3.1method",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_cac98fcd26c1",
+          "mean": 95,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p5 TableS1 entry3,scheme andsection3.1method",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        }
+      ],
+      "pairs": [
+        {
+          "item_i": "exp_41bd98dba742",
+          "item_j": "exp_484d7efaebd5",
+          "mean_difference": 8,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_41bd98dba742",
+          "item_j": "exp_389c830f7e89",
+          "mean_difference": 31,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_41bd98dba742",
+          "item_j": "exp_dd127eb2b1f5",
+          "mean_difference": 86,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_41bd98dba742",
+          "item_j": "exp_8545262bd121",
+          "mean_difference": 44,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_41bd98dba742",
+          "item_j": "exp_369ffbca2e28",
+          "mean_difference": 44,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_41bd98dba742",
+          "item_j": "exp_cac98fcd26c1",
+          "mean_difference": 2,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_484d7efaebd5",
+          "item_j": "exp_389c830f7e89",
+          "mean_difference": 23,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_484d7efaebd5",
+          "item_j": "exp_dd127eb2b1f5",
+          "mean_difference": 78,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_484d7efaebd5",
+          "item_j": "exp_8545262bd121",
+          "mean_difference": 36,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_484d7efaebd5",
+          "item_j": "exp_369ffbca2e28",
+          "mean_difference": 36,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_484d7efaebd5",
+          "item_j": "exp_cac98fcd26c1",
+          "mean_difference": -6,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_389c830f7e89",
+          "item_j": "exp_dd127eb2b1f5",
+          "mean_difference": 55,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_389c830f7e89",
+          "item_j": "exp_8545262bd121",
+          "mean_difference": 13,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_389c830f7e89",
+          "item_j": "exp_369ffbca2e28",
+          "mean_difference": 13,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_389c830f7e89",
+          "item_j": "exp_cac98fcd26c1",
+          "mean_difference": -29,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_dd127eb2b1f5",
+          "item_j": "exp_8545262bd121",
+          "mean_difference": -42,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_dd127eb2b1f5",
+          "item_j": "exp_369ffbca2e28",
+          "mean_difference": -42,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_dd127eb2b1f5",
+          "item_j": "exp_cac98fcd26c1",
+          "mean_difference": -84,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8545262bd121",
+          "item_j": "exp_369ffbca2e28",
+          "mean_difference": 0,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 0
+        },
+        {
+          "item_i": "exp_8545262bd121",
+          "item_j": "exp_cac98fcd26c1",
+          "mean_difference": -42,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_369ffbca2e28",
+          "item_j": "exp_cac98fcd26c1",
+          "mean_difference": -42,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        }
+      ],
+      "comparable_pairs": 20,
+      "quality_status": "passed"
+    },
+    {
+      "group_id": "g2",
+      "items": [
+        {
+          "item_id": "exp_066e33c896b9",
+          "mean": 94,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p9 TableS6 entry7,reaction/catalyststructures andsection3.3procedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_fba92a14e6bb",
+          "mean": 48,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p9 TableS6 entry2,reaction/catalyststructures andsection3.3procedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_9279aa7cd389",
+          "mean": 76,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p9 TableS6 entry6,reaction/catalyststructures andsection3.3procedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_8e5983854cd1",
+          "mean": 5,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p9 TableS6 entry5,reaction/catalyststructures andsection3.3procedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_343b2ac1c9a7",
+          "mean": 38,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p9 TableS6 entry3,reaction/catalyststructures andsection3.3procedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_e01ea673fa5d",
+          "mean": 90,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p9 TableS6 entry8,reaction/catalyststructures andsection3.3procedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_ea300526f709",
+          "mean": 44,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p9 TableS6 entry1,reaction/catalyststructures andsection3.3procedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        }
+      ],
+      "pairs": [
+        {
+          "item_i": "exp_066e33c896b9",
+          "item_j": "exp_fba92a14e6bb",
+          "mean_difference": 46,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_066e33c896b9",
+          "item_j": "exp_9279aa7cd389",
+          "mean_difference": 18,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_066e33c896b9",
+          "item_j": "exp_8e5983854cd1",
+          "mean_difference": 89,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_066e33c896b9",
+          "item_j": "exp_343b2ac1c9a7",
+          "mean_difference": 56,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_066e33c896b9",
+          "item_j": "exp_e01ea673fa5d",
+          "mean_difference": 4,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_066e33c896b9",
+          "item_j": "exp_ea300526f709",
+          "mean_difference": 50,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_fba92a14e6bb",
+          "item_j": "exp_9279aa7cd389",
+          "mean_difference": -28,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_fba92a14e6bb",
+          "item_j": "exp_8e5983854cd1",
+          "mean_difference": 43,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_fba92a14e6bb",
+          "item_j": "exp_343b2ac1c9a7",
+          "mean_difference": 10,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_fba92a14e6bb",
+          "item_j": "exp_e01ea673fa5d",
+          "mean_difference": -42,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_fba92a14e6bb",
+          "item_j": "exp_ea300526f709",
+          "mean_difference": 4,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_9279aa7cd389",
+          "item_j": "exp_8e5983854cd1",
+          "mean_difference": 71,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_9279aa7cd389",
+          "item_j": "exp_343b2ac1c9a7",
+          "mean_difference": 38,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_9279aa7cd389",
+          "item_j": "exp_e01ea673fa5d",
+          "mean_difference": -14,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_9279aa7cd389",
+          "item_j": "exp_ea300526f709",
+          "mean_difference": 32,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8e5983854cd1",
+          "item_j": "exp_343b2ac1c9a7",
+          "mean_difference": -33,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8e5983854cd1",
+          "item_j": "exp_e01ea673fa5d",
+          "mean_difference": -85,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_8e5983854cd1",
+          "item_j": "exp_ea300526f709",
+          "mean_difference": -39,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_343b2ac1c9a7",
+          "item_j": "exp_e01ea673fa5d",
+          "mean_difference": -52,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_343b2ac1c9a7",
+          "item_j": "exp_ea300526f709",
+          "mean_difference": -6,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        },
+        {
+          "item_i": "exp_e01ea673fa5d",
+          "item_j": "exp_ea300526f709",
+          "mean_difference": 46,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        }
+      ],
+      "comparable_pairs": 21,
+      "quality_status": "passed"
+    },
+    {
+      "group_id": "g3",
+      "items": [
+        {
+          "item_id": "exp_529d2754aa6a",
+          "mean": 93,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p12 FigureS2 scheme andcompleteprocedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        },
+        {
+          "item_id": "exp_94c775c85236",
+          "mean": 90,
+          "n": null,
+          "uncertainty_status": "unknown",
+          "uncertainty_scale": null,
+          "uncertainty_basis": "No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.",
+          "quality_status": "retained_with_flag",
+          "warnings": [
+            "Independent-mean uncertainty is unknown."
+          ],
+          "source_location": "Original-SI.pdf physical p12 FigureS2 scheme andcompleteprocedure",
+          "aggregation": "Published endpoint value without reweighting or selection of readings."
+        }
+      ],
+      "pairs": [
+        {
+          "item_i": "exp_529d2754aa6a",
+          "item_j": "exp_94c775c85236",
+          "mean_difference": 3,
+          "difference_se": null,
+          "uncertainty_status": "unknown",
+          "comparison_basis": "reported_value",
+          "comparable": 1
+        }
+      ],
+      "comparable_pairs": 1,
+      "quality_status": "passed"
+    }
+  ]
+}

tasks/alkene-transfer-conditions-ranking/tests/contract.json

@@ -0,0 +1,111 @@
+{
+  "task_id": "alkene-transfer-conditions-ranking",
+  "task_version": "1.0.0",
+  "groups": [
+    {
+      "group_id": "g1",
+      "label": "4-Phenyl-1-butene cross-metathesis adduct yield",
+      "item_ids": [
+        "exp_41bd98dba742",
+        "exp_484d7efaebd5",
+        "exp_389c830f7e89",
+        "exp_dd127eb2b1f5",
+        "exp_8545262bd121",
+        "exp_369ffbca2e28",
+        "exp_cac98fcd26c1"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    },
+    {
+      "group_id": "g2",
+      "label": "Rotenone cross-metathesis adduct yield",
+      "item_ids": [
+        "exp_066e33c896b9",
+        "exp_fba92a14e6bb",
+        "exp_9279aa7cd389",
+        "exp_8e5983854cd1",
+        "exp_343b2ac1c9a7",
+        "exp_e01ea673fa5d",
+        "exp_ea300526f709"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    },
+    {
+      "group_id": "g3",
+      "label": "Combined alkene yield from E/Z latent precursors",
+      "item_ids": [
+        "exp_529d2754aa6a",
+        "exp_94c775c85236"
+      ],
+      "direction": "ascending",
+      "group_weight": 0.3333333333333333,
+      "k": 2,
+      "epsilon": 0,
+      "uncertainty_model": "optional_reported_uncertainty",
+      "difference_standard_error": "sem_sum_upper_bound",
+      "endpoint": {
+        "name": "product 1H NMR yield",
+        "unit": "%",
+        "definition_en": "Reported1H NMR yield of the group-specified product relative to initial alkene or latent precursor,using methyl benzoate internal standard."
+      }
+    }
+  ],
+  "scoring_method": "spearman",
+  "primary_score": "task_spearman",
+  "spearman_range": [
+    -1,
+    1
+  ],
+  "spearman_tie_method": "average",
+  "spearman_undefined_policy": "task_error",
+  "spearman_aggregation": "Pearson correlation of true average ranks and submitted positions, separately within each group, then fixed group-weighted sum; do not pool groups or clip negative groups.",
+  "secondary_scoring_method": "equal_pair_inversion_random95",
+  "ranking_reward_range": [
+    0,
+    1
+  ],
+  "random_quantile": 0.95,
+  "unknown_uncertainty_policy": "use_reported_value",
+  "pair_rule": "For each unordered pair, count it iff abs(mu_i-mu_j)>k*s_ij+epsilon for known uncertainty, or >epsilon for unknown uncertainty. Each counted pair has equal weight within its group. Unknown uncertainty is not zero uncertainty.",
+  "ranking_reward_formula": "Validate finite 0.5<=q<1 and 0<=A<=1; if A==1 return 1; else if A<=q+1e-12 return 0; otherwise clamp((A-q)/(1-q),0,1). A is the fixed weighted sum of group pair accuracies; q is the whole-task random 95th percentile.",
+  "format_errors": {
+    "whole_task": "Core=-1; auxiliary=0 for malformed/oversized/non-UTF8 JSON, duplicate object keys, wrong/missing identity, invalid top-level/rankings, unknown or unassignable group.",
+    "known_group": "Missing/duplicate known group or invalid item permutation: raw group Spearman=null, aggregation value=-1, group pair accuracy=0; preserve group weight and apply auxiliary threshold after aggregation."
+  },
+  "max_submission_bytes": 1048576,
+  "minimum_conditions_per_group": 2,
+  "time_limit_sec": 28800,
+  "solver_resources": {
+    "cpus": 8,
+    "memory_mb": 16384,
+    "gpus": 1,
+    "gpu_types": [
+      "H20"
+    ]
+  },
+  "submission_path": "/logs/artifacts/rankings.json",
+  "report_path": "/logs/artifacts/report.md",
+  "sources_path": "/logs/artifacts/sources.json",
+  "analysis_path": "/logs/artifacts/analysis/",
+  "forbidden_result_scope": "Measured outcomes, yields, rankings, equivalent endpoints and result-fitted predictors for alkene transfer-reagent and catalyst conditions in the stated material and reaction system are prohibited."
+}

tasks/alkene-transfer-conditions-ranking/tests/grade.py

@@ -0,0 +1,174 @@
+"""Pipeline v18 Spearman core and preserved equal-pair auxiliary grader."""
+import argparse
+import copy
+import hashlib
+import json
+import math
+from pathlib import Path
+
+import original_grade as old
+
+TaskError = old.TaskError
+
+
+def bounded(value):
+    if not math.isfinite(value) or not -1 - 1e-12 <= value <= 1 + 1e-12:
+        raise TaskError('Spearman outside its finite range')
+    return min(1.0, max(-1.0, value))
+
+
+def average_ranks(values):
+    if len(values) < 2 or any(isinstance(x, bool) or not isinstance(x, (int, float))
+                              or not math.isfinite(x) for x in values):
+        raise TaskError('Invalid measurements')
+    ordered = sorted(range(len(values)), key=values.__getitem__)
+    ranks = [0.0] * len(values)
+    start = 0
+    while start < len(values):
+        stop = start + 1
+        while stop < len(values) and values[ordered[stop]] == values[ordered[start]]:
+            stop += 1
+        mean_rank = (start + 1 + stop) / 2
+        for index in ordered[start:stop]:
+            ranks[index] = mean_rank
+        start = stop
+    return ranks
+
+
+def spearman(values, predicted_positions):
+    ranks = average_ranks(values)
+    n = len(values)
+    if sorted(predicted_positions) != list(range(n)):
+        raise ValueError('Not a complete permutation')
+    predicted_ranks = [p + 1 for p in predicted_positions]
+    true_mean = math.fsum(ranks) / n
+    predicted_mean = math.fsum(predicted_ranks) / n
+    x = [r - true_mean for r in ranks]
+    y = [p - predicted_mean for p in predicted_ranks]
+    denominator = math.sqrt(math.fsum(r * r for r in x) * math.fsum(p * p for p in y))
+    if denominator <= 0:
+        raise TaskError('Undefined Spearman for constant results')
+    return bounded(math.fsum(r * p for r, p in zip(x, y)) / denominator)
+
+
+def promote_contract(contract):
+    value = copy.deepcopy(contract)
+    value.update(scoring_method='spearman', primary_score='task_spearman',
+                 spearman_range=[-1, 1], spearman_tie_method='average',
+                 spearman_undefined_policy='task_error',
+                 secondary_scoring_method='equal_pair_inversion_random95',
+                 ranking_reward_range=[0, 1])
+    return value
+
+
+def auxiliary_view(contract):
+    expected = {'scoring_method': 'spearman', 'primary_score': 'task_spearman',
+                'spearman_range': [-1, 1], 'spearman_tie_method': 'average',
+                'spearman_undefined_policy': 'task_error',
+                'secondary_scoring_method': 'equal_pair_inversion_random95',
+                'ranking_reward_range': [0, 1]}
+    if any(contract.get(k) != v for k, v in expected.items()):
+        raise TaskError('Invalid dual-score contract')
+    value = copy.deepcopy(contract)
+    value['scoring_method'] = value.pop('secondary_scoring_method')
+    for key in ('primary_score', 'spearman_range', 'spearman_tie_method',
+                'spearman_undefined_policy', 'ranking_reward_range'):
+        value.pop(key)
+    return value
+
+
+def scoring_bindings(contract, answers):
+    return {'contract_sha256': old.canonical_sha(contract),
+            'answers_sha256': old.canonical_sha(answers),
+            'grader_sha256': hashlib.sha256(Path(__file__).read_bytes()).hexdigest(),
+            'auxiliary_grader_sha256': hashlib.sha256(Path(old.__file__).read_bytes()).hexdigest()}
+
+
+def validate(contract, answers, baseline):
+    auxiliary = auxiliary_view(contract)
+    old.validate_task(auxiliary, answers)
+    if baseline.get('bindings') != scoring_bindings(contract, answers):
+        raise TaskError('Dual-score baseline hash mismatch')
+    old_baseline = {'q': baseline.get('q'), 'bindings': old.scoring_bindings(auxiliary, answers)}
+    old.validate_baseline(auxiliary, answers, old_baseline)
+    hidden = {g['group_id']: {i['item_id']: i['mean'] for i in g['items']} for g in answers['groups']}
+    for g in contract['groups']:
+        spearman([hidden[g['group_id']][i] for i in g['item_ids']], list(range(len(g['item_ids']))))
+    return auxiliary, old_baseline, hidden
+
+
+def add_core(contract, report, hidden):
+    report['scoring_method'] = 'spearman'
+    report['primary_score'] = 'task_spearman'
+    report['secondary_scoring_method'] = 'equal_pair_inversion_random95'
+    report['official_reward'] = None
+    report['official_ranking_reward'] = None
+    if not report['groups']:
+        report['task_spearman'] = -1.0
+        report['core_penalty_reason'] = 'whole_task_format_error'
+        report['groups'] = [{'group_id': g['group_id'], 'group_weight': g['group_weight'],
+                             'group_spearman_g': None, 'spearman_for_aggregation': -1.0,
+                             'group_pair_accuracy': 0.0, 'status': 'whole_task_format_error'}
+                            for g in contract['groups']]
+        return report
+    for g in report['groups']:
+        if g['status'] != 'valid':
+            rho = None
+        else:
+            item_ids = next(c['item_ids'] for c in contract['groups'] if c['group_id'] == g['group_id'])
+            positions = {item_id: i for i, item_id in enumerate(g['order'])}
+            rho = spearman([hidden[g['group_id']][i] for i in item_ids], [positions[i] for i in item_ids])
+        g['group_spearman_g'] = rho
+        g['spearman_for_aggregation'] = -1.0 if rho is None else rho
+    report['task_spearman'] = bounded(math.fsum(g['group_weight'] * g['spearman_for_aggregation']
+                                               for g in report['groups']))
+    return report
+
+
+def grade_submission(contract, answers, submission, baseline):
+    auxiliary, old_baseline, hidden = validate(contract, answers, baseline)
+    report = old.grade_submission(auxiliary, answers, submission, old_baseline)
+    report['random_q95'] = baseline['q']
+    return add_core(contract, report, hidden)
+
+
+def grade_file(contract, answers, path, baseline):
+    auxiliary, old_baseline, hidden = validate(contract, answers, baseline)
+    return add_core(contract, old.grade_file(auxiliary, answers, path, old_baseline), hidden)
+
+
+def official_scores(report, compliance, execution, memory=False):
+    if compliance == 'cheating':
+        return -1.0, 0.0
+    if compliance != 'compliant' or execution != 'completed':
+        return None, None
+    return report['task_spearman'], report['ranking_reward']
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--interface', choices=['txt', 'json'], default='txt')
+    parser.add_argument('--root', type=Path, default=Path('/tests'))
+    parser.add_argument('--submission', type=Path, default=Path('/logs/artifacts/rankings.json'))
+    parser.add_argument('--output', type=Path, default=Path('/logs/verifier'))
+    args = parser.parse_args()
+    args.output.mkdir(parents=True, exist_ok=True)
+    for name in ('reward.txt', 'reward.json', 'report.json', 'evaluator_error.json'):
+        (args.output / name).unlink(missing_ok=True)
+    try:
+        contract, answers, baseline = [old.load_json((args.root / f'{name}.json').read_bytes())
+                                      for name in ('contract', 'answers', 'random_baseline')]
+        report = grade_file(contract, answers, args.submission, baseline)
+    except (TaskError, old.SubmissionError, OSError, ValueError, KeyError, TypeError) as error:
+        old.dump(args.output / 'evaluator_error.json', {'status': 'task_error', 'error': str(error)})
+        raise SystemExit(2) from error
+    old.dump(args.output / 'report.json', report)
+    if args.interface == 'txt':
+        (args.output / 'reward.txt').write_text(str(report['task_spearman']) + '\n')
+    else:
+        old.dump(args.output / 'reward.json', {'reward': report['task_spearman'],
+                                              'ranking_reward': report['ranking_reward']})
+
+
+if __name__ == '__main__':
+    main()

tasks/alkene-transfer-conditions-ranking/tests/measurements.csv

@@ -0,0 +1,17 @@
+group_id,item_id,mean,n,uncertainty_status,uncertainty_scale,uncertainty_basis,quality_status,warnings,source_location,aggregation
+g1,exp_41bd98dba742,97,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p6 TableS3 entry2,catalyst structures;p5section3.1method",Published endpoint value without reweighting or selection of readings.
+g1,exp_484d7efaebd5,89,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p5 TableS1 entry2,scheme andsection3.1method",Published endpoint value without reweighting or selection of readings.
+g1,exp_389c830f7e89,66,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p5 TableS1 entry1,scheme andsection3.1method",Published endpoint value without reweighting or selection of readings.
+g1,exp_dd127eb2b1f5,11,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p6 TableS3 entry3,catalyst structures;p5section3.1method",Published endpoint value without reweighting or selection of readings.
+g1,exp_8545262bd121,53,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p6 TableS4 entry3,scheme;p5section3.1method",Published endpoint value without reweighting or selection of readings.
+g1,exp_369ffbca2e28,53,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p6 TableS4 entry2,scheme;p5section3.1method",Published endpoint value without reweighting or selection of readings.
+g1,exp_cac98fcd26c1,95,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p5 TableS1 entry3,scheme andsection3.1method",Published endpoint value without reweighting or selection of readings.
+g2,exp_066e33c896b9,94,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p9 TableS6 entry7,reaction/catalyststructures andsection3.3procedure",Published endpoint value without reweighting or selection of readings.
+g2,exp_fba92a14e6bb,48,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p9 TableS6 entry2,reaction/catalyststructures andsection3.3procedure",Published endpoint value without reweighting or selection of readings.
+g2,exp_9279aa7cd389,76,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p9 TableS6 entry6,reaction/catalyststructures andsection3.3procedure",Published endpoint value without reweighting or selection of readings.
+g2,exp_8e5983854cd1,5,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p9 TableS6 entry5,reaction/catalyststructures andsection3.3procedure",Published endpoint value without reweighting or selection of readings.
+g2,exp_343b2ac1c9a7,38,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p9 TableS6 entry3,reaction/catalyststructures andsection3.3procedure",Published endpoint value without reweighting or selection of readings.
+g2,exp_e01ea673fa5d,90,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p9 TableS6 entry8,reaction/catalyststructures andsection3.3procedure",Published endpoint value without reweighting or selection of readings.
+g2,exp_ea300526f709,44,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]","Original-SI.pdf physical p9 TableS6 entry1,reaction/catalyststructures andsection3.3procedure",Published endpoint value without reweighting or selection of readings.
+g3,exp_529d2754aa6a,93,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]",Original-SI.pdf physical p12 FigureS2 scheme andcompleteprocedure,Published endpoint value without reweighting or selection of readings.
+g3,exp_94c775c85236,90,,unknown,,No reliable independent-mean uncertainty is supplied. Unknown is not treated as zero.,retained_with_flag,"[""Independent-mean uncertainty is unknown.""]",Original-SI.pdf physical p12 FigureS2 scheme andcompleteprocedure,Published endpoint value without reweighting or selection of readings.

tasks/alkene-transfer-conditions-ranking/tests/original_grade.py

@@ -0,0 +1,423 @@
+"""Pipeline v16: equal pairs within groups; one task-level random-95 threshold."""
+
+import argparse
+import errno
+import hashlib
+import itertools
+import json
+import math
+import os
+from pathlib import Path
+import stat
+
+
+class TaskError(ValueError):
+    pass
+
+
+class SubmissionError(ValueError):
+    pass
+
+
+def canonical_sha(value):
+    return hashlib.sha256(json.dumps(value, sort_keys=True, separators=(",", ":"),
+                                     ensure_ascii=False, allow_nan=False).encode()).hexdigest()
+
+
+def scoring_bindings(contract, answers):
+    return {
+        "contract_sha256": canonical_sha(contract),
+        "answers_sha256": canonical_sha(answers),
+        "relations_sha256": canonical_sha([{"group_id": g["group_id"], "pairs": g["pairs"]}
+                                           for g in answers["groups"]]),
+        "weights_sha256": canonical_sha([{k: g[k] for k in ("group_id", "group_weight")}
+                                         for g in contract["groups"]]),
+        "grader_sha256": hashlib.sha256(Path(__file__).read_bytes()).hexdigest(),
+    }
+
+
+def validate_baseline(contract, answers, baseline):
+    if not isinstance(baseline, dict):
+        raise TaskError("random baseline must be an object")
+    q = finite(baseline.get("q"), "random_q95", 0.5)
+    if q >= 1:
+        raise TaskError("random_q95 must be below 1")
+    if contract.get("random_quantile") != 0.95:
+        raise TaskError("invalid random quantile")
+    if baseline.get("bindings") != scoring_bindings(contract, answers):
+        raise TaskError("random baseline hash mismatch")
+    return q
+
+
+def reward_from_accuracy(accuracy, q):
+    accuracy = finite(accuracy, "accuracy", 0)
+    q = finite(q, "random_q95", 0.5)
+    if accuracy > 1 or q >= 1:
+        raise TaskError("invalid accuracy or random threshold")
+    if accuracy == 1:
+        return 1.0
+    if accuracy <= q + 1e-12:
+        return 0.0
+    return min(1.0, max(0.0, (accuracy - q) / (1 - q)))
+
+
+def aggregate_accuracy(details):
+    # Preserve exact endpoints despite representational roundoff in weights.
+    if all(g["group_pair_accuracy"] == 1 for g in details):
+        return 1.0
+    return math.fsum(g["group_pair_accuracy"] * g["group_weight"] for g in details)
+
+
+def unique_object(pairs):
+    result = {}
+    for key, value in pairs:
+        if key in result:
+            raise SubmissionError("duplicate_json_key")
+        result[key] = value
+    return result
+
+
+def load_json(raw):
+    def reject_constant(value):
+        raise SubmissionError("nonfinite_json_constant")
+    return json.loads(raw.decode("utf-8"), object_pairs_hook=unique_object,
+                      parse_constant=reject_constant)
+
+
+def finite(value, name, minimum=None):
+    if isinstance(value, bool) or not isinstance(value, (int, float)):
+        raise TaskError(name + " is not numeric")
+    if not math.isfinite(value) or (minimum is not None and value < minimum):
+        raise TaskError(name + " is out of range")
+    return float(value)
+
+
+def difference_se(sem_i, sem_j, method):
+    if method == "sem_sum_upper_bound":
+        return sem_i + sem_j
+    if method == "independent_sem_quadrature":
+        return math.hypot(sem_i, sem_j)
+    raise TaskError("unsupported difference standard error method")
+
+
+def validate_task(contract, answers):
+    if not isinstance(contract, dict) or not isinstance(answers, dict):
+        raise TaskError("contract and answers must be objects")
+    if contract.get("scoring_method") != "equal_pair_inversion_random95":
+        raise TaskError("unsupported scoring method")
+    if contract.get("unknown_uncertainty_policy") != "use_reported_value":
+        raise TaskError("unsupported unknown-uncertainty policy")
+    for key in ("task_id", "task_version"):
+        if not isinstance(contract.get(key), str) or answers.get(key) != contract[key]:
+            raise TaskError("invalid task identity")
+    limit = contract.get("max_submission_bytes")
+    if type(limit) is not int or not 0 < limit <= 1048576:
+        raise TaskError("invalid submission size limit")
+    minimum_conditions = contract.get("minimum_conditions_per_group", 2)
+    if type(minimum_conditions) is not int or minimum_conditions < 2:
+        raise TaskError("invalid minimum conditions")
+    groups = contract.get("groups")
+    hidden_groups = answers.get("groups")
+    if not isinstance(groups, list) or not groups or not isinstance(hidden_groups, list):
+        raise TaskError("missing groups")
+    if len(groups) != len(hidden_groups):
+        raise TaskError("hidden group count mismatch")
+    hidden_map = {}
+    for hidden in hidden_groups:
+        if not isinstance(hidden, dict) or not isinstance(hidden.get("group_id"), str):
+            raise TaskError("invalid hidden group")
+        if hidden["group_id"] in hidden_map:
+            raise TaskError("duplicate hidden group")
+        hidden_map[hidden["group_id"]] = hidden
+    validated, group_ids, all_ids = [], set(), set()
+    for group in groups:
+        if not isinstance(group, dict):
+            raise TaskError("invalid contract group")
+        gid = group.get("group_id")
+        if not isinstance(gid, str) or not gid or gid in group_ids:
+            raise TaskError("invalid group id")
+        group_ids.add(gid)
+        ids = group.get("item_ids")
+        if (not isinstance(ids, list) or len(ids) < minimum_conditions
+                or any(not isinstance(item, str) or not item for item in ids)
+                or len(ids) != len(set(ids)) or all_ids.intersection(ids)):
+            raise TaskError("invalid item ids")
+        all_ids.update(ids)
+        if group.get("direction") != "ascending":
+            raise TaskError("only ascending order is supported")
+        group_weight = finite(group.get("group_weight"), "group_weight", 0)
+        if group_weight == 0:
+            raise TaskError("zero group weight")
+        k = finite(group.get("k"), "k", 0)
+        epsilon = finite(group.get("epsilon"), "epsilon", 0)
+        method = group.get("difference_standard_error")
+        if method not in ("sem_sum_upper_bound", "independent_sem_quadrature", "unknown"):
+            raise TaskError("unsupported difference standard error method")
+        if group.get("uncertainty_model") not in ("optional_reported_uncertainty",
+                "reported_sd_with_source_bounds", "independent_replicate_mean", "replicate_sd_upper_envelope", "unknown_replicate_hierarchy"):
+            raise TaskError("unsupported marginal uncertainty model")
+        hidden = hidden_map.get(gid)
+        if hidden is None or hidden.get("quality_status") != "passed":
+            raise TaskError("missing or invalid hidden group")
+        items = hidden.get("items")
+        if not isinstance(items, list) or len(items) != len(ids):
+            raise TaskError("hidden item count mismatch")
+        item_map = {}
+        for item in items:
+            if not isinstance(item, dict) or item.get("item_id") not in ids:
+                raise TaskError("invalid hidden item")
+            item_id = item["item_id"]
+            if item_id in item_map or item.get("quality_status") not in ("retained", "retained_with_flag"):
+                raise TaskError("duplicate or invalid hidden item")
+            mean = finite(item.get("mean"), "mean")
+            n = item.get("n")
+            if n is not None and (type(n) is not int or n < 1):
+                raise TaskError("invalid experimental replicate count")
+            uncertainty_status = item.get("uncertainty_status")
+            if uncertainty_status not in ("known", "unknown"):
+                raise TaskError("invalid uncertainty status")
+            uncertainty_model = group.get("uncertainty_model", "independent_replicate_mean")
+            if uncertainty_model == "reported_sd_with_source_bounds":
+                lo = finite(item.get("mean_lower"), "mean_lower")
+                hi = finite(item.get("mean_upper"), "mean_upper")
+                if lo > hi or not lo - 1e-12 <= mean <= hi + 1e-12:
+                    raise TaskError("inconsistent reported mean interval")
+            if uncertainty_status == "unknown":
+                if item.get("uncertainty_scale") is not None:
+                    raise TaskError("unknown uncertainty must be null")
+                sem = None
+            elif uncertainty_model == "optional_reported_uncertainty":
+                sem = finite(item.get("uncertainty_scale"), "uncertainty_scale", 0)
+                if not item.get("uncertainty_basis"):
+                    raise TaskError("known uncertainty needs a stated source basis")
+            elif uncertainty_model == "reported_sd_with_source_bounds":
+                if n is None or n < 2:
+                    raise TaskError("reported SD needs repeat evidence")
+                lower_sd = finite(item.get("sd_lower"), "sd_lower", 0)
+                upper_sd = finite(item.get("sd_upper"), "sd_upper", 0)
+                if lower_sd > upper_sd:
+                    raise TaskError("inconsistent reported SD interval")
+                sem = finite(item.get("uncertainty_scale"), "uncertainty_scale", 0)
+                expected_scale = upper_sd + (hi - lo) / 2
+                if not math.isclose(sem, expected_scale, rel_tol=1e-12, abs_tol=1e-12):
+                    raise TaskError("inconsistent source-bound uncertainty")
+            elif uncertainty_model in ("independent_replicate_mean", "replicate_sd_upper_envelope"):
+                sem = finite(item.get("uncertainty_scale"), "uncertainty_scale", 0)
+                sd = finite(item.get("sample_sd"), "sample_sd", 0)
+                if n is None or n < 2:
+                    raise TaskError("known SD needs repeat evidence")
+                expected_sem = sd / math.sqrt(n) if uncertainty_model == "independent_replicate_mean" else sd
+                if not math.isclose(sem, expected_sem, rel_tol=1e-12, abs_tol=1e-12):
+                    raise TaskError("inconsistent hidden uncertainty")
+            else:
+                raise TaskError("unsupported marginal uncertainty model")
+            item_map[item_id] = {"mean": mean, "sem": sem, "uncertainty_status": uncertainty_status}
+        stored = hidden.get("pairs")
+        if not isinstance(stored, list) or len(stored) != len(ids) * (len(ids) - 1) // 2:
+            raise TaskError("invalid hidden pair count")
+        stored_map = {}
+        for pair in stored:
+            if not isinstance(pair, dict):
+                raise TaskError("invalid hidden pair")
+            key = (pair.get("item_i"), pair.get("item_j"))
+            if key in stored_map:
+                raise TaskError("duplicate hidden pair")
+            stored_map[key] = pair
+        pairs = []
+        for item_i, item_j in itertools.combinations(ids, 2):
+            left, right = item_map[item_i], item_map[item_j]
+            difference = left["mean"] - right["mean"]
+            known = left["uncertainty_status"] == right["uncertainty_status"] == "known"
+            se = difference_se(left["sem"], right["sem"], method) if known else None
+            tolerance = k * se + epsilon if known else epsilon
+            comparable = int(abs(difference) > tolerance)
+            basis = "uncertainty_adjusted" if known else "reported_value"
+            if not math.isfinite(difference) or not math.isfinite(tolerance) or (known and not math.isfinite(se)):
+                raise TaskError("nonfinite pair arithmetic")
+            pair = stored_map.get((item_i, item_j))
+            if pair is None:
+                raise TaskError("missing hidden pair")
+            for name, expected, minimum in (("mean_difference", difference, None),):
+                actual = finite(pair.get(name), name, minimum)
+                if not math.isclose(actual, expected, rel_tol=1e-12, abs_tol=1e-10):
+                    raise TaskError("inconsistent hidden " + name)
+            if known:
+                if not math.isclose(finite(pair.get("difference_se"), "difference_se", 0), se,
+                                    rel_tol=1e-12, abs_tol=1e-10):
+                    raise TaskError("inconsistent pair uncertainty")
+            elif pair.get("difference_se") is not None:
+                raise TaskError("unknown pair uncertainty must be null")
+            if type(pair.get("comparable")) is not int or pair["comparable"] != comparable:
+                raise TaskError("inconsistent comparable flag")
+            if pair.get("uncertainty_status") != ("known" if known else "unknown"):
+                raise TaskError("inconsistent pair uncertainty status")
+            if pair.get("comparison_basis") != basis:
+                raise TaskError("inconsistent comparison basis")
+            pairs.append({"item_i": item_i, "item_j": item_j,
+                          "mean_difference": difference, "difference_se": se,
+                          "uncertainty_status": "known" if known else "unknown",
+                          "comparison_basis": basis, "comparable": comparable})
+        total = sum(pair["comparable"] for pair in pairs)
+        if total <= 0:
+            raise TaskError("group has no scoreable pair under the declared rule")
+        if type(hidden.get("comparable_pairs")) is not int or hidden["comparable_pairs"] != total:
+            raise TaskError("inconsistent comparable pair count")
+        validated.append({"group_id": gid, "item_ids": ids, "group_weight": group_weight,
+                          "pairs": pairs, "comparable_pairs": total})
+    if set(hidden_map) != group_ids:
+        raise TaskError("unknown hidden group")
+    if not math.isclose(math.fsum(g["group_weight"] for g in validated), 1.0,
+                        rel_tol=0, abs_tol=1e-12):
+        raise TaskError("group weights must sum to one")
+    return validated
+
+
+def zero_report(contract, reason):
+    return {"task_id": contract["task_id"], "task_version": contract["task_version"],
+            "ranking_reward": 0.0, "submission_status": reason, "groups": [],
+            "task_pair_accuracy": 0.0, "aggregation_zero_due_to_format": True,
+            "official_reward": None, "review_status": "not_reviewed", "evaluation_status": "completed"}
+
+
+def grade_submission(contract, answers, submission, baseline):
+    groups = validate_task(contract, answers)
+    q = validate_baseline(contract, answers, baseline)
+    if not isinstance(submission, dict):
+        return zero_report(contract, "invalid_top_level_structure")
+    if any(submission.get(key) != contract[key] for key in ("task_id", "task_version")):
+        return zero_report(contract, "task_identity_mismatch")
+    entries = submission.get("rankings")
+    if not isinstance(entries, list):
+        return zero_report(contract, "rankings_not_array")
+    known = {g["group_id"] for g in groups}
+    by_group = {}
+    for entry in entries:
+        if not isinstance(entry, dict) or not isinstance(entry.get("group_id"), str):
+            return zero_report(contract, "unassignable_group_entry")
+        if entry["group_id"] not in known:
+            return zero_report(contract, "unknown_group_id")
+        by_group.setdefault(entry["group_id"], []).append(entry)
+    details = []
+    for group in groups:
+        gid = group["group_id"]
+        matches = by_group.get(gid, [])
+        item = {"group_id": gid, "group_weight": group["group_weight"],
+                "group_pair_accuracy": 0.0, "status": "missing_group", "comparable_pairs": group["comparable_pairs"],
+                "total_pairs": len(group["pairs"]), "item_count": len(group["item_ids"]),
+                "coverage": group["comparable_pairs"] / len(group["pairs"])}
+        if len(matches) > 1:
+            item["status"] = "duplicate_group"
+        elif len(matches) == 1:
+            entry = matches[0]
+            order = entry.get("order")
+            if (not isinstance(order, list) or any(not isinstance(x, str) for x in order)
+                  or len(order) != len(group["item_ids"])
+                  or len(order) != len(set(order)) or set(order) != set(group["item_ids"])):
+                item["status"] = "invalid_permutation"
+            else:
+                positions = {item_id: i for i, item_id in enumerate(order)}
+                pair_details = []
+                for pair in group["pairs"]:
+                    wrong = ((positions[pair["item_i"]] - positions[pair["item_j"]])
+                             * pair["mean_difference"]) < 0
+                    pair_details.append({**pair, "reversed": wrong,
+                                         "counted_inversion": int(wrong and pair["comparable"] == 1)})
+                lost = sum(pair["counted_inversion"] for pair in pair_details)
+                total = group["comparable_pairs"]
+                reward = 1.0 - lost / total
+                item.update(status="valid", group_pair_accuracy=max(0.0, min(1.0, reward)),
+                            order=order, inversion_pairs=lost, correct_pairs=total-lost,
+                            comparable_pairs=total, total_pairs=len(pair_details),
+                            coverage=total/len(pair_details),
+                            reported_value_counted_pairs=sum(p["comparable"] for p in pair_details
+                                                             if p["comparison_basis"] == "reported_value"),
+                            uncertainty_adjusted_counted_pairs=sum(p["comparable"] for p in pair_details
+                                                                  if p["comparison_basis"] == "uncertainty_adjusted"),
+                            all_reported_value_unequal_pairs=sum(p["mean_difference"] != 0 for p in pair_details),
+                            all_reported_value_correct_pairs=sum(p["mean_difference"] != 0 and not p["reversed"]
+                                                                 for p in pair_details),
+                            pairs=pair_details)
+        details.append(item)
+    return {"task_id": contract["task_id"], "task_version": contract["task_version"],
+            "task_pair_accuracy": aggregate_accuracy(details),
+            "random_q95": q,
+            "ranking_reward": reward_from_accuracy(aggregate_accuracy(details), q),
+            "submission_status": "valid" if all(g["status"] == "valid" for g in details) else "group_errors",
+            "groups": details, "official_reward": None,
+            "review_status": "not_reviewed", "evaluation_status": "completed"}
+
+
+def read_regular(path, limit):
+    try:
+        fd = os.open(path, os.O_RDONLY | os.O_NOFOLLOW | os.O_NONBLOCK)
+        try:
+            info = os.fstat(fd)
+            if not stat.S_ISREG(info.st_mode):
+                raise SubmissionError("not_regular_file")
+            if info.st_size > limit:
+                raise SubmissionError("submission_too_large")
+            with os.fdopen(fd, "rb", closefd=False) as stream:
+                raw = stream.read(limit + 1)
+            if len(raw) > limit:
+                raise SubmissionError("submission_too_large")
+            return raw
+        finally:
+            os.close(fd)
+    except OSError as error:
+        if error.errno in (errno.ENOENT, errno.ELOOP, errno.EISDIR, errno.ENOTDIR,
+                           errno.ENXIO, errno.EACCES, errno.EPERM):
+            raise SubmissionError("missing_or_unreadable_submission") from error
+        raise
+
+
+def grade_file(contract, answers, path, baseline):
+    validate_task(contract, answers)
+    q = validate_baseline(contract, answers, baseline)
+    try:
+        raw = read_regular(path, contract["max_submission_bytes"])
+        submission = load_json(raw)
+    except (SubmissionError, UnicodeError, json.JSONDecodeError, RecursionError) as error:
+        report = zero_report(contract, str(error) or type(error).__name__)
+        report["random_q95"] = q
+        return report
+    report = grade_submission(contract, answers, submission, baseline)
+    report["random_q95"] = q
+    report["submission_sha256"] = hashlib.sha256(raw).hexdigest()
+    return report
+
+
+def dump(path, value):
+    path.write_text(json.dumps(value, indent=2, allow_nan=False) + "\n", encoding="utf-8")
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--contract", type=Path, default=Path("/tests/contract.json"))
+    parser.add_argument("--answers", type=Path, default=Path("/tests/answers.json"))
+    parser.add_argument("--baseline", type=Path, default=Path("/tests/random_baseline.json"))
+    parser.add_argument("--submission", type=Path, default=Path("/logs/artifacts/rankings.json"))
+    parser.add_argument("--output-dir", type=Path, default=Path("/logs/verifier"))
+    args = parser.parse_args()
+    args.output_dir.mkdir(parents=True, exist_ok=True)
+    for name in ("reward.txt", "reward.json", "report.json", "evaluator_error.json"):
+        (args.output_dir / name).unlink(missing_ok=True)
+    try:
+        contract = load_json(args.contract.read_bytes())
+        answers = load_json(args.answers.read_bytes())
+        baseline = load_json(args.baseline.read_bytes())
+        report = grade_file(contract, answers, args.submission, baseline)
+    except (TaskError, SubmissionError, OSError, ValueError, KeyError, TypeError) as error:
+        dump(args.output_dir / "evaluator_error.json", {"status": "task_error", "error": str(error)})
+        raise SystemExit(2) from error
+    dump(args.output_dir / "report.json", report)
+    (args.output_dir / "reward.txt").write_text(str(report["ranking_reward"]) + "\n", encoding="utf-8")
+    print(json.dumps({"ranking_reward": report["ranking_reward"],
+                      "submission_status": report["submission_status"],
+                      "task_pair_accuracy": report["task_pair_accuracy"],
+                      "random_q95": report["random_q95"],
+                      "aggregation_zero_due_to_format": report.get("aggregation_zero_due_to_format", False),
+                      "official_reward": None, "review_status": "not_reviewed", "evaluation_status": "completed"}))
+
+
+if __name__ == "__main__":
+    main()